You Needn’t be the First Investor There:

First-Mover Disadvantages in Emerging Private Equity Markets

Alexander Peter Groh*

This version: 15 September 2016

Abstract: Early movers into emerging private equity markets may capitalize on growth opportunities and build-up networks. However, the deal-making conditions might not (yet) be favorable for the pioneers and they could lack local experience. I support this latter conjecture by analyzing a unique hand-collected dataset on emerging private equity market transactions. From 1,157 deals in 86 host countries between 1973 and 2009 I find that “waiting and learning pays”. Controlling for opportunity cost of capital, cross currency rate fluctuations, GDP growth over the holding periods, for liquidity of the exit markets, and for socio-economic factors, such as a country’s innovation capacity, its legal quality, its human capital and labor market frictions, the analyses reveal that early transactions underperform the later ones. I interpret this as a result from learning benefits and from improvements of the deal-making environment over time. The learning benefits are stronger if investors are located in the country of the investee firm.

JEL codes: G23, G24, O16, P52

Keywords: Emerging Markets, Venture Capital, Private Equity, Alternative Assets

*) EMLYON Business School, 23 Avenue Guy de Collongue, 69132 Ecully, France groh@em- lyon.com

1

Electronic copy available at: http://ssrn.com/abstract=2839215 You Needn’t be the First Investor There:

First-Mover Disadvantages in Emerging Private Equity Markets

This version: 15 September 2016

Abstract: Early movers into emerging private equity markets may capitalize on growth opportunities and build-up networks. However, the deal-making conditions might not (yet) be favorable for the pioneers and they could lack local experience. I support this latter conjecture by analyzing a unique hand-collected dataset on emerging private equity market transactions. From 1,157 deals in 86 host countries between 1973 and 2009 I find that “waiting and learning pays”. Controlling for opportunity cost of capital, cross currency rate fluctuations, GDP growth over the holding periods, for liquidity of the exit markets, and for socio-economic factors, such as a country’s innovation capacity, its legal quality, its human capital and labor market frictions, the analyses reveal that early transactions underperform the later ones. I interpret this as a result from learning benefits and from improvements of the deal-making environment over time. The learning benefits are stronger if investors are located in the country of the investee firm.

JEL codes: G23, G24, O16, P52

Keywords: Emerging Markets, Venture Capital, Private Equity, Alternative Assets

2

Electronic copy available at: http://ssrn.com/abstract=2839215 By entering into a ‘new market’ firms can exploit significant growth opportunities without generating excess capacity (Spence, 1979). Empirical research supports the notion that there are important advantages to being the first entrant in some sorts of markets (Schmalensee, 1982).

First-mover advantages are expected for the pioneering firms with respect to gaining a head-start over rivals. This opportunity may occur because the firm possesses some unique resources or foresight (Lieberman and Montgomery, 1988). A first-mover advantage may also result from learning. Spence (1981) demonstrates that if learning can be kept proprietary, the learning curve can generate substantial barriers to entry. However, the mechanisms that benefit the first-mover may be counterbalanced by disadvantages, such as the ability of followers to free ride on the first- movers’ activities.

Emerging private equity (PE) markets are “new markets” in the investors’ universe. First- movers might be able to benefit from untapped deal flow and gain a head-start over rivals. The emerging countries’ economic growth requires substantial funding. Early movers could capitalize on the growth opportunities, thereby achieving local experience and building up networks. For investors in emerging market PE funds there is a widespread argument to commit capital early:

Access to the top general partners might be restricted after they proof a successful track record.

Their follow-on funds will be oversubscribed and therefore not accessible by limited partners who had no relationship with them in the pioneering phase. This motivates institutional investors to move quickly into emerging regions and to develop strategic partnerships with local fund managers.

Nevertheless, a supportive environment for PE deal making does not only require economic growth. “Traditional” PE markets (e.g. the U.S., the U.K., Canada, France, Japan, Germany etc.)

3

do not exhibit GDP growth rates as high as the average emerging country but have many other favorable characteristics. These characteristics are subject of empirical studies and include the depth of a country’s capital market (Jeng and Wells, 2000, Black and Gilson, 1998, and Gompers and Lerner 2000), labor market frictions, investor protection, general ease of doing business, a country’s human capital (Lazear, 1990, Blanchard, 1997, La Porta et al., 1997 and 1998, Roe,

2006, Glaeser et al., 2001, Djankov et al., 2003 and 2008, Lerner and Schoar,2005, Cumming et al., 2006 and 2010), and its innovation capacity (Gompers and Lerner, 1998 and Kortum and

Lerner, 2000) among others.

Requiring these favorable conditions for successful PE deal making, it might not be easy for emerging market pioneers to gain first-mover advantages because the conditions have not yet developed to the necessary extent. It seems rather beneficial to wait until the countries’ state of development facilitates PE deal-making.

Despite the required level of socio-economic development of a country for PE activity, research has provided evidence on the value drivers in PE transactions and on appropriate target company characteristics: It is commonly understood that PE sponsors can add value to their portfolio companies by reducing agency costs and by exploiting debt tax shields. Active monitoring, incentive structuring, and bonding may lower the agency costs of free cash flow.

Improvements in operating efficiency, focusing on core activities and selling off non-core assets creates more competitive entities (Jensen, 1988, Kaplan, 1989a and 1989b, Lichtenberg and

Siegel, 1990, Muscarella and Vetsuypens, 1990, Smith, 1990, Wruck, 1990 and Kaplan, 1991).

The benefit of PE is assumed to be largest for corporations with high agency cost of free cash flow, namely high operating cash flows but missing investment opportunities. Jensen (1986)

4

notes that “…desirable leveraged buyout candidates are frequently firms or divisions of larger firms that have stable business histories and substantial free cash flow (i.e., low growth prospects and high potential for generating cash flows) - situations where agency costs of free cash flow are likely to be high.”

I argue that the required socio-economic environment for successful PE deal-making has not yet established in many emerging PE markets, but especially not at the time of the pioneers’ entrants. I also expect that creating deal flow is cumbersome because potential target firms do not fulfill the appropriate characteristics of ideal buyout candidates with potential high agency costs of free cash flow. It is doubtful that companies in rapidly growing markets provide these opportunities because they have high capital expenditures. The necessity to exploit the growth potential in a short time should also not leave much room for operational improvement. Further, as the opportunity cost of capital is usually high in emerging countries, other investment channels, such as portfolio and corporate foreign direct investment could be better suited to capitalize on the growth anticipations. The assumption of a lack of appropriate buyout target candidates is supported by the observation that many of the emerging market PE deals are indeed not “true buyouts” but rather disguised project, infrastructure or growth financing transactions.

As a result, we do not expect first mover advantages from early entry in emerging PE markets.

There is no need to be among the first investors because PE deal making conditions are not sufficiently developed at the time the pioneers enter. Furthermore, deal making is different in emerging countries. Investors need to gain local experience and should not expect that the

“traditional” PE deal model they are confident with can be easily transferred to emerging environments.

5

I support this claim by analyzing the largest dataset from Lopez-de-Silanes et al. (2015) on emerging PE markets used so far.1 I find from 1,157 emerging market PE transactions in 86 host countries between 1973 and 2009 that “waiting and learning pays”. I control for different definitions of the opportunity cost of capital, for cross currency rate fluctuations, for real GDP growth during the transaction holding period, for liquidity of the exit market, and for socio- economic determinants, such as a country’s innovation capacity, its legal quality, its human capital and labor market protection and find that early transactions underperform the later ones.

The performance of the PE transactions (measured by their internal rates of return), increases over time since the pioneering investment. This effect is consistent with benefits from learning and improvements of PE deal making conditions. However, learning effects are not proprietary but benefit the whole investment community because later entrants also earn higher returns. I therefore conclude that there are no directly measurable first mover advantages and that it is preferable for investors to delay their emerging market PE allocations until the particular countries are mature enough to establish liquid PE activity. The data also reveals that if investors aim to enter in emerging PE markets, then they should allocate to general partners located in the host country of investments. Cross border deals significantly yield lower returns than national transactions and distantly managed funds (e.g. those headquartered in the U.S.) underperform their peers which are closer to the investment host country. The findings are robust with respect to a variety of potentially confounding factors. I show that cross currency rate changes, alternative measures for the opportunity cost of capital, eventually over-represented countries in our sample, or misspecifications of the dates of pioneering transactions does not affect the

1 I am very grateful to Ludovic Phalippou for making the data available for this paper.

6

findings. I also reveal that the learning benefit decays over time so that emerging PE market performance turns to an average level.

The paper is structured as follows: I first discuss related literature and then develop the hypotheses in the following session. Subsequently, I describe the data, perform the analyses and robustness checks and interpret the results before I conclude.

1. Related Literature

While there are several contributions focusing on the performance of PE funds and leveraged buyout transactions in general, research on emerging market PE returns is limited. Only a few publications include transactions in developing countries in their sample – and if, they usually focus on general investment activity but not on returns to investors.

Strömberg (2008) provides a comprehensive overview of global LBO activity. He collects data of 21,397 LBO transactions between 1970 and 2007 with a total transaction value of approximately $ 3.6 trillion from a commercial provider and documents the tremendous growth of the industry but especially of the geographic dispersion. Nevertheless, the LBO transactions outside North America and Western Europe are still relatively few and account for only approximately 13% of the global volume in numbers and 7% in value.

Kaplan and Stömberg (2009) note that PE activity strongly spread to new parts of the world between 2001 and 2006, particular to Asia, where deal size almost tripled during that period.

Lerner and Schoar (2004) argue that systematic data on emerging market PE returns is hard to come by but submit that returns in these nations appear to have been far lower than in the U.S.

7

and Europe. They conclude that the experience of PE funds in the developing world poses interesting issues which have been little explored in academic research so far.

Lerner and Schoar (2005) address the contractual structures of emerging market PE transactions. In low enforcement and civil law countries, PE sponsors tend to use common stock and straight debt and therefore, rely on equity and board control. This might alleviate enforcement problems which could result from contractual provisions linked to convertible preferred securities which are commonly used in the U.S. At the same time, the transactions in inferior enforcement countries have lower valuations and returns. Kaplan et al. (2007) provide detailed analyses of contracts used for venture capital transactions in 23 countries and compare them with the typical U.S. style contract at that time. They find that contracts differ strongly across legal regimes and conclude that transactions using non-U.S. style contracts fail more often.

Leeds and Sunderland (2003) underscore the notion of inferior returns to investors from emerging market PE activities and discuss potential determinants of under-performance. They argue that the PE industry evolved gradually in the U.S. over a forty year period which was increasingly conducive to this type of financing and point to a sympathetic public policy environment, a reliable legal system, stability, a well-developed financial market, and finally, demand from cooperative entrepreneurs. In particular, they identify low standards of corporate governance, limited legal recourse, and dysfunctional capital markets impeding PE activity in emerging markets. Nahata et al. (2014) elaborate on these assumed deficiencies and reveal that indeed, the quality of legal rights and investor protection, and the general development of stock markets are inhibitors. Cumming and Walz (2009) and Cao et al. (2014) confirm the role of legal protection for PE sponsors in emerging countries. Lopez-de-Silanes et al. (2015) find that

8

emerging market PE transactions have slightly longer durations and exhibit statistically significantly poorer performance across several measures with the exception of bankruptcy rates.

However, the authors expected the opposite due to higher assumed cost of capital in these countries. The lower returns could be the result of costly learning, poor legal environments and illiquid exit markets. They also find lower degrees of leverage of developing country PE transactions and suggest this as another reason for smaller returns on equity. Demiroglu and

James (2010) find that LBO activity and the success likelihood is largely driven by debt market conditions and especially by loan-spreads. Ivashina and Kovner (2011) quantify the expected increase on the return on equity of a buyout sponsor caused by a decrease of the cost of debt and suggest that PE firms should repeatedly interact with banks to maintain relationships and offer cross selling opportunities to reduce the credit spread for their investee firms. Their finding that lower cost of debt increase the return on equity (all else equal) is consistent with standard theory on capital structuring decisions. Nevertheless, Axelson et al. (2013) cannot replicate the results of

Demiroglu and James (2010) and of Ivashina and Kovner (2011) with their sample of 1,157 buyouts in developed economies. Instead, they conclude that pricing of buyout transactions is cyclical and driven by the cost of debt. High deal leverage is associated with higher transaction prices and lower equity returns suggesting that acquirers overpay when access to debt is easy.

Nevertheless, access to debt financing is more restricted in emerging countries and cost of debt is generally higher. Therefore, the conclusions drawn in Lopez-de-Silanes (2015) seem to better explain emerging countries’ lower PE returns.

Taussig (2011) addresses the liability of foreignness of investors who are present in emerging market PE transactions. He argues that foreignness could lead to being discriminated against by local market participants and regulators or to being at an information disadvantage when

9

investing in emerging country target firms. Nevertheless, the setting changes at exit because then foreign PE investors are able to benefit from their international networks and reputations to sell stakes in companies from countries with weaker formal contracting institutions. In this context, positive institutional change during an investment’s holding period should reduce acquirers’ concerns to buy from only locally originating PE firms, thus reducing the competitive advantage of foreign PE funds. Hence, returns on investment of internationally acting PE funds should be negatively affected by improvements of formal contracting institutions in low enforcement countries. Taussig (2011) receives support for this claim by running multivariate analyses on 267

PE transactions in emerging countries originated from OECD countries.

Chemmanur et al. (2014) add to the research on the liability of foreignness of PE investors.

They find that syndicates composed of international and local investors are more successful than syndicates of either exclusively international or exclusively local funds. Both groups of investors have comparative disadvantages: International PE firms lack proximity but local funds might have less investment experience. The benefits of mixed syndicates are stronger in emerging regions which is consistent with the notion that difficulties in monitoring and deficiencies in local knowledge faced by international investors are more severe there.

Reddy and Blenman (2015) analyze LBO transactions in different growth phases of the investees and compare developed and developing economies. They find that the returns achieved by the financial sponsors are on average higher for transactions in the developed countries.

However, in periods of strong economic growth, the returns are higher in the emerging economies.

10

This paper contributes to the existing literature by focusing on the socio-economic determinants of the internal rates of return of emerging market PE transactions. I thereby use the most comprehensive and accurate data set on emerging market PE transactions ever used so far in academic research. This allows replicating some of the findings above but provides additional evidence on the fact that there is a timing effect in emerging PE markets. Using a detailed mapping of the entry order in every particular country covered by the data set I reveal that pioneers do not earn superior returns. Instead the performance of emerging market PE transactions increases over time. I argue that this increase in returns to investors can be attributed to learning and improvements of the deal-making environment. The learning speed is higher for

GPs who are located inside the target country and not abroad. However, learning cannot be kept proprietary because later entrants also earn higher returns with the time elapsing since a pioneer’s entry. Therefore, followers benefit from waiting, while the deal making environments improve. I conclude that it is not necessary to enter early into untapped PE markets.

2. Hypotheses development

The pioneering emerging market PE investor is exposed to particular risks and deal making constraints. Probably, there is lacking deal supporting infrastructure, such as investment banks, law firms, consultants and accountants specialized on the typical issues of PE transactions.

Furthermore, investment managers of the pioneering funds might have substantial experience in advanced PE markets, such as North America or Western Europe. However, in geographic and cultural distant locations, this knowledge could not be of high value because of doing business differences. Additionally, managers and employees of local investee firms might not have experience either with the new type of investor and governance structure. Therefore, all market

11

participants need to gain experience. I hypothesize that the more the time elapses, the higher the success of emerging market PE transactions due to learning effects. This contradicts the hypothesis of first-mover advantages: If there were first-mover advantages, e.g. because the pioneers get access to untapped superior quality deal flow, the earlier transactions should be more successful than the later ones. I further hypothesize that building up emerging market PE experience requires local presence. Differences in the deal making conditions between developed and emerging markets might be too large to gain experience without proximity to the investee firms. Making deals from distant locations or across borders therefore diminishes learning effects.

In parallel to the first hypothesis, I assume that PE is a financial intermediary relationship which requires a level of development not necessarily reached in many of the emerging countries.

Certain socio-economic characteristics need to be present for a successful establishment of a PE market. The farther these characteristics advance the more successful PE deal making becomes.

Emerging countries typically improve these characteristics during their socio-economic development. However, once again, this takes time and I therefore hypothesize that PE deal making is not only depending on learning, investor or investee specific determinants but also on the socio-economic environment in the emerging host country. The higher the development level of this environment the more successful become PE transactions.

Despite eventual less favorable deal making conditions in emerging compared to developed countries the developing markets have one major advantage: their expected economic catch-up potential. The most important rationale for emerging market PE investments is the anticipation of strong GDP growth much above that of developed countries. The high economic growth rates

12

may even compensate for deal making obstacles, such as missing deal supporting infrastructure and not yet mature socio-economic conditions. In any case, we hypothesize that economic growth rates strongly affect the success of emerging market PE transactions.

We finally assume that even if economic growth is one of the expected drivers of success, the deals still underlie the typical market conditions. These conditions are valuation multiples, financing cost and exit market liquidity. I therefore hypothesize that not only emerging market experience, GDP growth and socio-economic factors shape returns to investors. There are additional determinants such as stock market valuations, cost of debt and exit market liquidity which affect successful PE deal making.

I address these hypotheses as subsequently described.

3. Data Set and Descriptive Statistics

A. General Sample Characteristics and Dependent Variable

The sample is based on 1,655 emerging market PE transactions collected from Private

Placement Memoranda (PPMs) and extracted from the data used in Lopez-de-Silanes et al.

(2015). The dataset offers a broad pool of variables describing each single investment.

Unfortunately, some data records are only partly complete and thus, several transactions need to be dropped.

The most important information for our purpose is the timing of the transaction, the host country of the target and its success expressed by the internal rate of return (IRR) of the underlying cash flow stream. The IRRs are gathered gross of management fees and are therefore comparable across time, countries and general partners. Unfortunately, not all of the transactions

13

are exited at the time of the creation of the dataset. Therefore, I have to rely on the reported net asset values to assess their IRRs. This imposes uncertainty with respect to the real returns finally earned in these transactions. The proportion of reported net asset values in our sample naturally falls with older funds’ vintage years. As a consequence, I observe more reported net asset values for the deals at the end of the sampling period. Brown et al. (2014) and Jenkinson et al. (2013) find that reported net asset values are generally conservative. Hence, any bias induced by reported net asset values works towards my finding that later emerging market PE transactions outperform the earlier ones.

Since this paper focuses on “emerging” PE markets I pay particular attention on the fact that several of the sample countries have developed to an “advanced” state during the observation period. I refer to the IMF definition of emerging and advanced economies and further discard transactions which have been closed after the host country changed its status from “emerging” to

“advanced”. This reduces the number of “real” emerging market PE transactions to 1,157.2 The sample comprises investments in 86 host countries by 73 different general partners between 1973 and 2009 with exit/reporting dates ranging from 1975 to 2009 and durations from one month up to 18.5 years. Table 1 presents these primary sample characteristics.

======

Insert Table 1 here

======

2 In more detail, I discard the transactions in the Czech Republic with a closing date after 2008, in Hong Kong after 1997, in Singapore after 1997, in Slovakia after 2008, in Slovenia after 2006, in South Korea after 1997 and in Taiwan after 1997.

14

The table reveals the representativeness of our sample in terms of geography, closing and exit timing, and transaction duration. Nevertheless, it seems that some countries, e.g. Poland or South

Africa are over-represented. This potential over-representation will be addressed in robustness checks. The timing information for the individual deals is given in monthly accuracy suggesting to use these months’ ends as settlement dates for all benchmark comparisons. However, for some transactions, no information on the closing or exit month is given. For these cases, I use 30 June of the same year as the reference date for all benchmark calculations. Figure 1 presents the distribution of the transaction closing dates in the sample. It reveals that the bulk of the transactions were made after 1987. However, several deals were closed earlier, but exclusively in

Hong Kong.

======

Insert Figure 1 here

======

The origins of the investors are not as broadly diversified as the host countries of the investees. This corresponds with the typical pattern of emerging market PE transactions being originated in financial centers. Almost 50% of our sample transactions are sponsored by general partners based in the US. 11% are undertaken by UK GPs. Other hubs are in Poland, Finland, the

Netherlands, the Czech Republic, the Russian Federation, and in Greece serving the Central and

Southern Eastern European and the Commonwealth of Independent States PE markets (for approximately 16% of our sample transactions). GPs from financial centers in China, India, Hong

Kong, and Malaysia provide financing for 10% of the transactions mainly located in South

Eastern Asia. South Africa serves as a hub for 108 African transactions and Argentina and Brazil

15

for 49 Latin American deals which represents approximately 9%, respectively 4% of the sample.

From this geographic distribution of investor locations I realize that 73% of the transactions are sourced across a border while 27% are local deals, i.e. emerging country investors investing in the same country’s investees.

The paper focuses on the development of PE returns across countries and over time.

Therefore, the most important information is the annualized gross internal rate or return (Gross

IRR) of every individual transaction as disclosed in the PPM. It constitutes the dependent variable throughout most of our analyses. The IRR is calculated from a USD investor’s point of view in most of the cases in our PPMs. However, in several of the transactions the IRR has been calculated in EUR, GPB, YPN, or ZAR. I convert these returns into USD IRRs by correcting for the cross currency rate changes between closing and exit of the transactions.

A few transactions yielded extraordinary high returns, beyond conventional levels. To correct for these outliers, I winsorize the IRR distribution at the 95-percentile and present it in Figure 2.

Table 2 is a breakdown of these IRRs (in USD) according to the host countries of investment.

======

Insert Figure 2 and Table 2 here

======

The mean (winsorized) IRR of the 1,157 transactions is 18.8% while their median is 15.1%.

The upper limit of the (winsorized) IRRs is 148% while some transactions wiped out the invested capital. The right-skewed distribution is typical for returns in the PE asset class. Mean comparisons test reveal at high levels of significance (all levels below 0.01) that cross border deals yield 13% lower IRRs than national deals, that US sourced transactions return 23% less vs

16

sourcing in other countries and that earlier transactions (split at the median level of entry in the sample) yield a 10% lower IRR than the subsequent deals.

The industry classifications of the sample transactions follow Fama and French (1997). I identify 47 different industries. Out of all investments, approximately 14% were made in companies in the Trading and 11% in the Communications industries. The remainder of the sample is broadly diversified among the industry spectrum. For 11 transactions, I do not have information about the investee firms’ industry and treat them as others/unknown. The industry segmentation of the sample is presented in Table 3.

======

Insert Table 3 here

======

B. Independent and Control Variables

I collect several independent and control variables. Most importantly, the key variable of interest is a proxy for the experience that investors have in the various countries. Since I do not know detailed information about the sponsors at the time of the transactions, e.g. their staff’s emerging market experience or their tenure/experience in general I refer to simple proxies that can be determined from my own data collection. The first stylized measure for experience in a particular country uses our sample’s first transaction entry date: The pioneering observation in every investment host country sets the cut-off entry date and all subsequent transactions are related to this date. The key variable (1) “Time Since 1st Investment in the Host Country” takes the value of the difference of the closing dates between a particular transaction and the pioneering one in the same country. Obviously, for the initial transactions in every country this variable has

17

a value of zero. This proxy therefore captures the experience that the overall investment community has in an individual country. It can also be interpreted from another perspective: It likewise serves as a proxy for the development and state of the PE market and characterizes the awareness of the asset class in a country’s society and professional investment community.

However, the variable needs to be interpreted with some caution because I am unaware if the first observation is indeed the pioneering deal in a particular country. Therefore, the proxy might be biased towards too small values. I address this concern in robustness checks.

The second alternative experience measure focuses on the investors directly. Variable (2),

“GP's Experience in the Host Country”, assesses the knowledge a particular general partner (GP) has gained in the respective country, independent of all other market participants. It is defined similar to the primary key variable: The first transaction of a general partner in a particular country sets the offset date for this general partner in this respective country. All subsequent transactions of the same general partner in the same country are related to this date. The measure is the difference between the closing date of any of a general partner’s transactions and the closing date of its first deal in this country. Since every general partner can invest in several countries, the variable takes more often zero values than the primary key variable. However, this variable does not have a potential bias caused by eventually missing observations of the pioneering investment. GPs report their complete track records in PPMs. Thus, I should have collected the correct entry dates in a particular country for every GP.

======

Insert Table 4 here

======

18

Table 4 presents the key measures of experience and all other independent and control variables that I use in our multivariate analyses. Public stock markets have an important signaling effect for the unquoted equity market. PE investors usually use public peer group multiples when valuing unquoted investees. Therefore, stock market valuations strongly drive target values and, hence, the returns of PE transactions (Lopez de Silanes et al., 2015). However, it is not clear which stock markets serve to determine peer group valuations for emerging market PE transactions. Market participants could either refer to peers traded in the U.S. or locally. Variable

(3), the time-matching S&P 500 return, is therefore used to determine the benchmark return if valuations follow U.S. peers and variable (4), the time-matching local or regional stock market return, determines the public market equivalent return if peers are selected from local (emerging) stock markets. However, several of the investee host-countries do/did not (at transaction date) have a public stock market or a representative benchmark index. For these countries, I refer to close neighbors or regionally representative indices to determine the alternative benchmark returns. Table 5 lists all the emerging market benchmark indices that I consider and provides information how often they serve as benchmark and which alternative indices are used.

======

Insert Table 5 here

======

Variable (5), (of Table 4) the time-matching S&P 500 return in local currency, and variable

(6), the time-matching local or regional stock market return in local currency, are based on variables (3) and (4), but converted to benchmark returns in local (emerging market) currencies.

They are used in a robustness check which addresses if the detected effects result from currency

19

exchange rate fluctuations between USD and emerging country currencies over the transaction holding periods.

Variable (7), the time matching GDP growth, measures the real growth of the gross domestic product in the investment host-country over the holding period of the transaction. The GDP growth is annualized and extrapolated according to the exact (months’ ends) closing and exit days. Variable (8), the aggregated IPO proceeds in the investment host country in the year of exit, captures the liquidity of the exit market at the time of divestment. Several countries don’t have public stock markets/IPO activity and hence, consistently 0 proceeds. Variable (9), the host country’s global innovation index, is an indicator assessing the innovative capacity of countries.

Variable (10), a host country’s quality of the educational system controls for the available human capital in a country. Variable (11), a host country’s interest rate spread in the year of closing, is a proxy for access to debt and debt financing cost at origination of the transactions. Variable (12), the Difficulty of Firing Index, is a doing business indicator which assesses labor market frictions and the ability of investors to implement new strategies during their holding period. Variable

(13), a host country’s Property Rights Index, controls for legal quality in the investee country.

Variable (14) is a dummy signaling if a PE fund’s headquarter is in the U.S. and variable (15) is another dummy indicating a cross border transaction, i.e. the GP and investee firm are not located in the same country. The numbers of GPs in other fund locations in our sample are too small to further distinguish.

Variables (9), (10), (12), and (13) are often not available for years prior to 2000 and do not change meaningfully over time either. I therefore consider them time-invariant and use their 2009

20

values. All other variables either match the duration of the sample transactions or correspond with the entry or exit/reporting year observation.

Table 6 presents the descriptive statistics of our independent and control variables.

======

Insert Table 6 here

======

From Table 6, I realize that the mean and median IRR in local currencies are slightly higher than calculated in USD. This is analog for the variables measuring the public market equivalent returns in USD and in local currencies, i.e. variable (3) and (5) and variables (4) and (6). This signals, on average, a depreciation of the emerging market currencies against the USD. The relatively small public market equivalent returns of the S&P 500 index can be explained by the observation period. As presented in Figure 1, a bulk of our sample transactions were closed after the 2000 peak of the index and hence, actually have negative benchmark returns. The other variables exhibit rather intuitive distributions.

======

Insert Table 7 here

======

Table 7 shows the correlation matrix among all independent and control variables. It reveals a high correlation (0.61) between the two experience measures, and relatively high negative correlations (-0.3 and -0.42) between the S&P 500 benchmark returns (in USD) and the two

21

experience measures. The correlations are equivalently high (-0.41 and -0.31) if the S&P 500 benchmark returns are calculated in local currencies. Therefore, I use each experience measure exclusively (as one possible alternative) throughout the regression specifications and analyze the impact of the stock market indices in a separate robustness check. All other independent and control variables are introduced stepwise to rule out any concern about multicolinearity.

Unexpectedly, the time matching emerging stock market benchmark returns only correlate weakly with the S&P 500 public market equivalents (variable pairs [3] with [4] and [5] with [6]).

I have expected a stronger globalization effect among the public equity markets. The high correlations between (3) and (5) and between variables (4) and (6) reveal that cross-currency rate changes do not meaningfully affect the benchmark returns on average.

4. Multivariate Analyses

I address my hypotheses with several stepwise OLS regressions. The independent variable is always the winsorized IRR of the sample transactions either calculated in USD or in local currencies. All standard errors are robust.

a. PE Returns from the USD Investor Perspective

In the first specification in Table 8, Panel A I regress the winsorized IRR in USD on the experience measure “Time Since the 1st Investment in the Host Country”, a constant and controls including country, GP, industry fixed effects and the legal quality indicator. The number of observations is 1157, and the adjusted R2 is 22.74%.

22

The coefficient of the independent variable in the first line of Table 8 is 0.013 and statistically significant at the 1% level. The second line (0.119) represents the standardized parameter coefficient, i.e. the estimate if all variables are transferred into their z-scores. The third line

[0.005] is the estimate’s standard error.

======

Insert Table 8 here

======

The economic magnitude of the parameter estimate can be interpreted as follows: Every year of waiting until a transaction has been closed led to an increase of the IRR by 1.3 percentage points on average. This effect is strong and meaningful for investors and I attribute it to two factors: The first is the experience that market participants gained in the particular emerging country. The second is the simple result of further development of the deal making conditions that facilitate PE transactions in this country.

Specifications B to E add independent variables and alter the set of control variables. In specification B, I include the transaction period-matching S&P 500 benchmark return as independent variable. As expected, the global stock market benchmark has a strong impact on valuations in the non-quoted emerging markets and hence, on the PE returns: A 1% increase in the benchmark return over the holding period increases the transactions’ returns by 1.15%-points, on average. The standardized parameter coefficient (0.273) is slightly larger than that of our experience measure, signaling that the public stock market has a stronger effect on PE

23

performance than the experience measure. The adjusted R2 of this regression increases from

22.74% to 26.55% compared to specification A.

Specification C adds the transaction time-matching local benchmark return as independent variable. Its parameter 0.143 is significant at the 1%-level, indicating that investors do not only refer to the U.S. to determine valuation multiples, but also to local stock markets. Alternatively, one could argue that the local stock market indices provide more appropriate information about the local economic conditions than the S&P 500 index. However, the local benchmarks’ standardized coefficient is smaller than that of the other variables, suggesting that the local benchmark returns have less impact on emerging market PE returns than experience or the S&P

500 levels. It is furthermore surprising that both variables gain, respectively keep significance in this and the subsequent regressions and do not cancel out each other. I assumed that controlling for the standard benchmark index performance (i.e. the S&P 500 index) is sufficient to explain emerging market PE returns and therefore address the impact of the benchmark indices and potential multicolinearity in a separate robustness check.

From specification D onwards, country specific properties are included and therefore, country fixed effects are dropped to avoid multicolinearity. The first country specific characteristic is the holding period-matching real GDP growth. This yields a decrease of the adjusted R2 from

26.94% to 22.66%, compared to specification C. The regression reveals that the GDP growth’s parameter coefficient is significant at the 1% level. Economic growth is the underlying source of value creation in the investee corporations and therefore strongly drives the transactions’ returns.

The impact of GDP growth is substantial. A 1%-point increase of the real GDP growth improves the average transaction performance by 2.5%-points.

24

Specification E adds the IPO proceeds in the country in the transaction’s exit year to the set of independent variables. The coefficient for the proxy of exit market liquidity is significant at the

1% level and highlights its importance for the asset class.

Regression specification F of Panel B of Table 8 also includes the proxy for innovation capacity, and the dummy variable for the general partners’ location in the U.S. At the same time, the fund fixed effects are dropped because of their colinearity with the U.S. fund location dummy. The coefficient of the Global Innovation Index is significant at the 1% level and highlights the importance of technological innovations also for emerging market PE returns. At the same time, the fund location dummy has a strong and significant negative coefficient. Since this dummy variable also resembles a geographic distance parameter (all other transactions are managed at closer distances), this signals that locally (or closer to the investee countries) managed emerging market PE portfolios yielded higher returns. From regression F, I interpret that portfolios managed from the U.S. returned 9.9% less return to their investors on average, all else equal. The reference group is all other general partners, either located in emerging markets directly or more closer to them.

I use the alternative experience measure “GP’s Experience in the Host Country” in regressions

G and H which produce statistically and economically less powerful results, revealed by smaller standardized coefficients and larger standard errors. However, I interpret the consistently positive significant parameter coefficients as evidence of the learning effects that general partners have after the entry into a particular emerging country. They gain experience and benefit from delaying investments due to improving deal-making conditions. Nevertheless, the gained experience cannot be kept proprietary as demonstrated by the above analyses using the alternative

25

key variable “Time Since the 1st Investment in the Host Country” which measures the experience and waiting benefits to all participants. All players in emerging PE markets benefit from additional experience and improving deal making conditions.

Furthermore, I add the quality of education in the host country in specification G. The parameter’s coefficient is significant at the 10% level. This highlights the importance of human capital for the asset class. Countries with higher education quality are eventually better suited to emerge appropriate PE investee firms.

Regression H also includes the host country’s interest rate spread at transaction closing and its difficulty of firing index. Unfortunately, this reduces the number of observations by 49 because the indicators are both not available for a number of countries. Both parameters have significant negative coefficients. First, this provides evidence that the cost of debt affect PE returns in the expected direction for emerging markets. Second, the implementation of strategic changes is less likely to be successful in host countries with higher labor market frictions. However, inclusion of the two additional variables (or only one of them) cancels out the quality of the educational system indicator.

b. PE Returns from the Local Investor Perspective

PPMs as fundraising documents usually address the international investment community and are edited using USD as reference currency. This is also valid for my sample with a very few exceptions where the reference currency is EUR, GBP, YPN or ZAR. Whatever reference currency is chosen, the reported IRRs might be affected by exchange rate changes between the reference currency and the local (emerging country) currency. I recall that the value creation (or

26

destruction) happens at the level of the investee firm. Using foreign multiples to benchmark and foreign currencies to settle transactions imposes additional variability and a potential bias of the reported IRRs. Therefore, I calculate all IRRs and all detected parameters also from a local investor’s perspective and rerun the above presented regressions. Table 9 presents the four most important regressions using IRRs and the benchmark returns in local (emerging market) currencies.

======

Insert Table 9 here

======

Specification I from Table 9 regresses the transactions’ winsorized IRR in local (emerging market) currencies on the experience measure, on the time-matching S&P 500 and the time- matching local stock market return, both also converted into local currencies, on the controls for country, fund, industry and legal quality, and on a constant. Specification J adds time-matching

GDP growth, the countries’ aggregated IPO proceeds in the exit year, the global innovation index score and the GP headquarter dummy for the U.S. At the same time, country and GP fixed effects are dropped. Regressions K and L use the alternative experience measure and add the emerging countries’ quality of the educational system (in K) and its difficulty of firing index (in L). Again, including the difficulty of firing index reduces the number of observations to 1108.

The four regressions provide evidence that the detected implications are not affected by cross currency rate changes. All parameter coefficients retain their significance, their sign and their economic magnitude if the determinants are converted to local currencies. The most important

27

result is that the IRRs of the emerging market PE transactions did not improve over time due to appreciations of their countries’ currencies against the USD but because of the experience that investors gained respectively due to improving deal-making conditions.

c. GP Location and Learning Effects

Next, I analyze the impact of geographical and cultural distance in emerging market PE transactions on the learning curve. As hypothesized above, the positive effects from gaining experience in emerging market PE deal making could be smaller without local presence of the

GPs. The differences in doing PE business between the developed countries where most of the

GPs of the sample are located and the host countries of investment might be large and cause difficulties for investors to gain experience without local presence. Therefore, I introduce a proxy for geographic and cultural distance between the investor and the investee, a dummy variable which is equal to one if the GP and the target firm are not in the same country, i.e. if a particular deal is a cross border transaction.

When including this dummy in the regressions, I need to discard variables causing multicolinearity, i.e. GP fixed effects and host country fixed effects. The rationale is simple:

Several GPs exclusively do cross border deals and, similarly, many countries only receive PE funding from outside locations.

======

Insert Table 10 here

======

28

Panel A of Table 10 presents regressions equivalent to those presented in Table 8, Panel A, but adding a dummy variable which indicates cross border transactions and its interaction term with

“Time Since the 1st Investment in the Host Country”. The dependent variable is always the winsorized IRR of the PE transactions from a USD investor's point of view. Standard errors are robust. Specification M regresses the dependent variable on investee industry fixed effects, the legal quality indicator and the cross border deal dummy variable. It reveals that the IRRs of cross border deals are 11.6% below the transactions which are nationally originated. Specification N shows that the learning effect remains if we control for cross border transactions. In specification

O, I add the interaction term between the first two variables. The cross border dummy itself is no longer significant but the negative parameter of the interaction term provides evidence that the benefit from learning diminishes in cross border transactions. The remaining specifications support the robustness of the previous results.

In Panel B of Table 10, I use the alternative experience measure “GP’s Experience in the Host

County” and its interaction term with cross border transactions as main independent variables.

The regressions R to U reveal that GPs should stay local to benefit most from learning how to perform emerging market PE transactions. The economic and statistical significance of the interaction term is even stronger than if the key variable “Times since 1st Investment in the Host

Country” was used.

d. Robustness of the Results

The results could be affected by a non-representative sample of emerging market transactions or be exposed to correlating covariates. I run therefore a series of robustness checks to reveal that this is not the case.

29

I broke-down the number of observations of the sample per country in Table 1. I have no influence on the sample’s geographical distribution but use all data gained. It seems that South

Africa or Poland is over-represented in terms of the size of their economy, their population or with respect to their PE activity. Additionally, any other country with a large number of observations, e.g. Brazil or China, could bias the results. I partly control for a potential geographical bias via the inclusion of country fixed effects. However, in some of the regression specifications, I drop these controls. Therefore, I repeat regression F presented in Table 8 but always discard one of the countries which might be the source of a geographical bias. Table 10 presents the results of the regressions which address this potential bias, where we first individually drop South Africa, then Poland, followed by Brazil and China.

======

Insert Table 11 here

======

Table 11 provides evidence that the main results are not affected by a potential overweight of one particular country. Most of the parameter coefficients keep their economic and statistical significance if I drop a particular country from the sample. The only two exceptions are the dummy variable for the U.S. headquarter when South Africa is excluded from the observations and the liquidity of the IPO market if China is dropped. I consider the loss of significance of these two parameter coefficients as not affecting the main contribution of this paper and note that completely dropping the countries individually also induces a geographical bias.

30

In a subsequent step, I address the robustness of the results with respect to potential multicolinearity among covariates. I realize from Table 7 that (among the important independent variables) the benchmark returns correlate strongest with each other and with the primary variable of interest, the experience measure “Time Since the 1st Investment in the Host Country”.

Even if it seems off-key to assess returns of PE transactions in a regression model without controlling for a public market equivalent I drop the benchmark returns alternatively, and both of them together in a robustness check and show that the regressions still yield qualitatively the same results. I present the repetition of specification F (from Table 8) but drop either benchmark return in specifications V and W, and both of them in specification X of Table 12.

======

Insert Table 12 here

======

The regressions in Table 12 reveal that the significance of the experience measure decreases slightly in specifications V and X, which are both after dropping the S&P 500 benchmark returns.

In comparison with specification W, I interpret that the local or regional stock market benchmark returns have a smaller effect on the quality of the regression model and the results. Altogether, the principal results hold, even without consideration of a public market equivalent return.

In a final robustness check, I verify that the results are not exposed to a potential bias caused by the very early transactions in the sample or by not having captured the “really pioneering” transactions. As discussed above, and from Figure 1, I recall that the bulk of transactions were made from 1988 onwards. Nevertheless, 17 closings were made before 1988, but exclusively in

31

Hong Kong. These early transactions could bias the findings. Additionally, I cannot be certain that the sample includes the “really pioneering” investment in the particular host countries. There are no systematic records of early emerging market PE transactions and it is therefore difficult if not impossible to detect the PE transaction that was really the first one in a particular country.

Hence, I analyze the impact of selecting another date than our observed first closings as the offset to calculate the experience measure “Time Since the 1st Investment in the Host country”. If a later deal than the first one determines the offset and the results still hold, then the findings are stable over time and robust to missing the “real pioneering” deal. I therefore discard the first five

(arbitrarily chosen) transactions in every country from our sample. This creates new cut-off dates for the experience measure and drops 284 observations. I repeat the regressions from Table 8,

Panel A and present the robustness checks in Table 13.

======

Insert Table 13 here

======

The first regression specifications Y to AC in Table 13 provide evidence that the findings are not affected by a later cut-off date which determines the experience measure “Time since 1st

Investment in a Host Country”. The final column reveals that dropping the transactions prior to

1988 does likewise not affect our results. I conclude that the results are stable over some time and also robust to potential outliers caused by the very early PE activity in Hong Kong or by not observing the “really pioneering” investment.

32

e. Decay of the Benefit of Waiting

The results confirm that the accumulation of experience or simply waiting for improved deal- making conditions increases the IRR of emerging market PE transactions. However, this effect cannot sustain over time and needs to decay (otherwise, the transaction returns would continuously grow to infinity). The decline of the marginal benefit of waiting is best modelled in a level-log regression where the transactions’ IRR is regressed on the natural logarithm of the experience measure and additional independent and control variables. Using the log of the experience variable turns it into an elasticity measure and therefore also addresses the uncertainty about the real pioneering transaction in every country of our sample. Table 14 presents regressions AD to AF, where I first add one year to “Time Since the 1st Investment in the Host

Country” (for the pioneering transaction in each country) and then compute its natural logarithm.

All other variables are kept in levels.

======

Insert Table 14 here

======

The results presented in Table 14 provide evidence that the marginal positive effect of waiting to enter an emerging PE market on transaction IRRs declines over time. I interpret this finding consistent with the intuitive expectation that learning effects are limited and that appropriate deal- making conditions establish after several years. Lacking additional learning effects and with more favorable deal making environments emerging market PE returns tend to an average.

33

5. Discussion and Limitations

The analyses reveal a strong effect of waiting on the performance of emerging market PE transactions. The positive parameter coefficient of the elapsed time since the pioneering investment in a host country is economically and statistically significant throughout all of our regressions and robust with respect to the effect of alternative benchmark indices, cross currency rate changes, potential over-representation of sample countries, non-observation of the “really pioneering” investment, and to the fact that it cannot last forever. The measure for the particular general partners’ experience in the various countries follows the same pattern. I interpret this as evidence of learning benefits and of the improvement of the deal making conditions. However, as measured by the key variable “Time Since the 1st Investment in the Host Country”, these effects are not proprietary to a single general partner. All market participants benefit from gaining experience and improving deal making conditions at the same time. The benefit from learning is stronger if the general partners play locally, i.e. if they are located in the same country as their investees. The learning speed is lower in cross border transactions. I interpret this as support for the notion that it is necessary to gain deal making experience in the particular host countries because doing business conditions and cultural differences might not allow a simple transfer of

Western oriented PE deal models on emerging market transactions.

I aim to control for deal making conditions, using benchmark index performance, GDP growth, exit market liquidity and cost of debt at the same time of the transaction. However, I cannot completely cover the improving socio-economic conditions during the holding period over a particular deal. The reason is that the indicators for innovation capacity, the quality of educational systems or for labor market rigidities are sticky over time and not correctly tracking

34

gradual improvements. Furthermore, since these indicators are not available for many of the sample countries for the respective transaction periods, I use the 2009 observations of these indicators for all countries and treat them time-invariant. For this reason, I cannot completely disentangle learning effects from improving socio-economic conditions.

The analyses replicate previous findings, proving that several independent variables have the expected economic effects on PE performance. I confirm, for example, the impact of GDP growth, exit market and socio-economic conditions on the value generation in emerging market

PE transactions. I also show the importance of stock market valuations thereby differentiating a global standard (the S&P 500 index) and local benchmark indices. It is not unexpected that each of both measures for opportunity cost has an individual and alternative contribution to explain PE returns. However, it is rather surprising that both measures remain significant in the regressions if kept simultaneously. This highlights the notion that emerging market PE valuations follow global benchmarks but underlie specific conditions at the same time, which are well expressed by superimposed local emerging stock market conditions.

A potential limitation of this research is set by the dataset. It is possible that the sample of transactions does not include the “real first movers”. These might have been extraordinary successful transactions caused by privileged access to superior untapped deal flow of the pioneer.

In general, I believe to have collected the most comprehensive and accurate sample of (early) emerging market PE transactions. Nevertheless, I cannot rule out that I do not precisely track entry and early followers in every particular country. However, the institutional investors who provided PPMs for this research are prominent players, especially for emerging PE markets.

PPMs are marketing instruments for fundraising. If one of the general partners in the fundraising

35

process had originated one of the potentially extraordinary pioneering transactions I would expect it not to be missing in its PPM. Therefore, I would rather believe to miss a part of the early less successful transactions in the sample instead of the popular ones. Since in general, the coverage of transactions improves over time, this potential sample selection bias works towards the results and not against them.

Nevertheless, I might not have access to the GP who actually made the pioneering investment in a particular country and hence, the cut-off date for "Time Since the 1st Investment in the Host

Country” is flawed. This may cause bias in the cross sectional comparisons. I address such a potential bias by dropping the first five transactions in every country from the sample in a robustness check. This check reveals the general trend of increasing PE returns over time and proofs that this trend is independent of the first deal observation in the particular countries.

6. Conclusions

I address the question if investors should enter emerging PE markets at such a pace as it has been observed in the past. There are several good reasons to justify early entry. Many emerging countries are large in terms of their population and have enormous economic catch up potential.

This eventually results in high demand for risk capital and favorable economic growth for a foreseeable future. One may consider this an optimal environment for PE investors. However, sophisticated relationships between investors, general partners and investee firms require certain legal standards, enforcement opportunities and briefly, a socio-economic development level which might not yet be reached in many emerging countries, but was certainly not reached at the time of the pioneering PE investments. Additionally, in strongly growing economies deal flow of

“ideal investees” with high agency cost of free cash flows and appropriate debt capacity might

36

not emerge. It is therefore possible that investors in emerging PE markets focus on project, infrastructure or expansion financing opportunities instead of traditional buyouts. But, nevertheless, my analyses show that - be it traditional buyouts, or be it deals with project, infrastructure or expansion financing characteristics - the returns to investors from emerging market PE transactions gradually increased over time. Investors would have been better-off if they had waited with their allocations. This finding is strong and robust with respect to a variety of possible concerns. I interpret it as the result from learning and improving deal making conditions. I control for deal making conditions but the availability of data does not allow to completely disentangling the two effects.

Early entry might still be beneficial for investors to build up relationships with local general partners, allowing further allocations and increased returns in the future. This is a widely used argument in practice. It is valid if the experience the general partners gain increases their success rates at later stages. However, I also show that emerging PE market experience cannot be kept proprietary. All investors benefit from waiting, thus ruling out an important principle of gaining first-mover advantages.

Hence, I conclude that investors needn’t be the first ones to invest in emerging market PE.

However, more positive effects from early movers might materialize in fund vintages subsequent to the coverage of our sample – when the PE markets gain maturity. This conjecture may be addressed in additional research when performance figures of more recent vintage years become available.

37

References

Axelson, U., Jenkinson, T., Strömberg, P. and Weisbach, M. S. (2013): Borrow cheap, buy high? The determinants of leverage and Pricing in buyouts, Journal of Finance, Vol. 68, No. 6, pp. 2223-2267.

Black, B. and Gilson, R. (1998): Venture Capital and the structure of capital markets: Banks versus stock markets, Journal of Financial Economics, Vol. 47, No.3, pp. 243-277.

Blanchard, O. J. (1997): The medium run, Brookings Papers on Economic Activity, 1997, No. 2, pp. 89-158.

Brown, G. W., Gredil, O. and Kaplan, S. N. (2014): Do private equity funds game returns? Fama-Miller working paper, available at SSRN: http://ssrn.com/abstract=2271690.

Cao, J. X., Cumming, D., Qian, M. and Wang, X. (2014): Creditor rights and LBOs, Journal of Banking and Finance, forthcoming.

Chemmanur, T. J., Hull, T. J. and Krishnan, K. (2014): Do local and international venture capitalists play well together? A study of international venture capital investments, unpublished working paper, available at http://ssrn.com/abstract=1670319.

Cumming, D., Fleming, G. and Schwienbacher, A. (2006): Legality and venture capital exits. In: Journal of Corporate Finance, Vol. 12, pp. 214 – 245

Cumming, D. and Walz, U. (2009): Private equity returns and disclosure around the world, Journal of International Business Studies, Vol. 41, No. 4, pp. 727-754.

Cumming, D., Schmidt, D. and Walz, U. (2010): Legality and venture capital governance around the world, Journal of Business Venturing, Vol. 25, No. 1, pp. 54–72.

Demiroglu, C. and James, C. M. (2010): The role of private equity group reputation in LBO financing, Journal of Financial Economics, Vol. 96, pp. 306-330.

Djankov, S., La Porta, R., Lopez-de-Silanes, F. and Shleifer, A. (2003): Courts, Quarterly Journal of Economics, Vol. 118, No. 2, pp. 453-517.

Djankov, S., La Porta, R., Lopez-de-Silanes, F. and Shleifer, A. (2008): The law and economics of self-dealing, Journal of Financial Economics, Vol. 88, pp. 430–465.

38

Fama, E. and Kenneth, F. (1997): Industry cost of equity, Journal of Financial Economics, Vol. 43/2, pp. 153-193.

Glaeser, E. L., Johnson, S. and Shleifer, A. (2001): Coase vs. the Coasians, Quarterly Journal of Economics, Vol. 116, pp. 853-899.

Gompers, P. and Lerner, J. (1998): What drives venture fundraising? Brooking Papers on Economic Activity, Microeconomics, pp. 149-192.

Gompers, P. and Lerner, J. (2000): Money chasing deals? The impact of funds inflows on the valuation of private equity investments, Journal of Financial Economics, Vol. 55, No.2, pp. 281- 325.

Groh, A. P. and Liechtenstein, H. (2009): How attractive is CEE for risk capital investors?, Journal of International Money and Finance, Vol. 28, No. 4, pp. 625-647.

Ivashina, V. and Kovner, A. (2011): The private equity advantage: Leveraged buyout firms and relationship banking, Review of Financial Studies, Vol. 24, pp. 2462-2498.

Jenkinson, T., Sousa, M. and Stucke, R. (2013): How fair are the valuations of private equity funds?, unpublished working paper, available at http://ssrn.com/abstract=2229547.

Jensen, M. C. (1986): Agency costs of free cash flow, corporate finance, and takeovers, American Economic Review, Vol. 76, pp. 323-329.

Jensen, M. C. (1988): Takeovers: Their causes and consequences, Journal of Economic Perspectives, Vol. 2, pp. 21-48.

Jeng, L. A. and Wells, P. C. (2000): The deteminants of Venture Capital funding: Evidence across countries, Journal of Corporate Finance, Vol. 6, No. 3, pp. 241-289.

Kaplan, S. N. (1989a): The effects of management buyouts on operating performance and value, Journal of Financial Economics, Vol. 24, pp. 217-254.

Kaplan, S. N. (1989b): Management buyouts: Evidence on taxes as source of value, Journal of Finance, Vol. 44, pp. 611-632.

Kaplan, S. N. (1991): The staying power of leveraged buyouts, Journal of Financial Economics, Vol. 29, pp. 287-313.

39

Kaplan, S. N., Martel, F. and Strömberg, P. (2007): How do legal differences and experience affect financial contracts?, Journal of Financial Intermediation, Vol. 16, pp. 273-311.

Kaplan, S. N. and Strömberg, P. (2009): Leveraged buyouts and private equity, Journal of Economic Perspectives, Vol. 23, No. 1, pp. 121-146.

Kortum, S. and Lerner, J. (2000): Assessing the contribution of venture capital to innovation, Rand Journal Economics, Vol. 31, No. 4, pp. 674-692.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1997): Legal determinants of external finance, Journal of Finance, Vol. 52, No. 3, pp. 1131-1150.

La Porta, R., Lopez-de-Silanes, F., Shleifer, A. and Vishny, R. (1998): Law and finance, Journal of Political Economy, Vol. 106, No. 6, pp. 1113-1155.

Lazear, E. P. (1990): Job security provisions and employment, Quarterly Journal of Economics, Vol. 105, pp. 699-726.

Leeds, R. and Sunderland, J. (2003): Private equity in emerging markets: Rethinking the approach, Journal of Applied Corporate Finance, Vol. 15, pp. 111-119.

Lerner, J. and Schoar, A. (2004): Private equity in the developing world: The determinants of transaction structures, unpublished working paper.

Lerner J. and Schoar, A. (2005): Does legal enforcement affect financial transactions? The contractual channel in private equity, Quarterly Journal of Economics, Vol. 120, No. 1, pp. 223- 246.

Lichtenberg, F. R. and Siegel, D. (1990): The effects of leveraged buyouts on productivity and related aspects of firm behavior, Journal of Financial Economics, Vol. 27, pp. 165-194.

Lieberman, M. B. and Montgomery, D. B. (1988): First-mover advantages, Strategic Management Journal, Vol. 9, 41-58.

Lopez-de-Silanes, F., Phalippou, L. and Gottschalg, O. (2015): Giants at the gate: Diseconomies of scales in private equity, Journal of Financial and Quantitative Analysis, forthcoming.

40

Muscarella, C. J. and Vetsuypens, M. R. (1990): Efficiency and organizational structure: A study of reverse LBOs, Journal of Finance, Vol. 45, pp. 1389-1413.

Nahata, R., Hazarika, S. and Tandon, K. (2014): Success in global venture capital investing: Do institutional and cultural differences matter?, Journal of Financial and Quantitative Analysis, Vol. 49, No. 4, pp. 1039-1070.

Reddy, N. and Blenman, L. (2015): Leveraged buyout activity: A tale of developed and developing economies, Journal of Financial Markets, Money and Institutions, forthcoming.

Roe, M. (2006): Political determinants of corporate governance, Oxford

Schmalensee, R. (1982): Differentiation advantages of pioneering brands, American Economic Review, Vol. 72, No. 3, pp. 349-365.

Smith, A. J. (1990): Corporate ownership structure and performance: The case of management buyouts, Journal of Financial Economics, Vol. 27, pp. 143-164.

Spence, A. M. (1979): Investment strategy and growth in a new market, Bell Journal of Economics, Vol. 10, No. 1, pp. 1-19.

Spence, A. M. (1981): The learning curve and competition, Bell Journal of Economics, Vol. 12, No. 1, pp. 49-70.

Strömberg, P. (2008): The new demography of private equity, unpublished working paper, available at http://www.sifr.org/PDFs/stromberg(demography2008).pdf?q=predicting-ipo- failures-in-the-old-and-new-economies.

Taussig, M. (2011): Which firms benefit from strengthening contract enforcement? Foreignness as an asset at exit in emerging economies private equity, unpublished working paper, available at http://ssrn.com/abstract=1927028.

Wruck, K. H. (1990): Financial distress, reorganization, and organizational efficiency, Journal of Financial Economics, Vol. 27, pp. 419-444.

41

Table 1: Primary characteristics of our sample of transactions This table presents the countries where the target firms of our sample transactions are headquartered, the number of observations, the observations as a percentage of the total sample, the first and last observed acquisition closing and exit/reporting years, and the mean, median, minimum and maximum duration of the transactions. Obs. Closing Year Exit/Reporting Duration Year Country Obs. in % Min Max Min Max Mean Median Min Max Algeria 1 0.10 2003 2003 2007 2007 4.0 4.0 4.0 4.0 Angola 2 0.20 2004 2005 2008 2008 4.0 4.0 3.5 4.5 Argentina 40 3.50 1992 2008 1998 2008 8.9 8.8 0.5 16.5 Bangladesh 1 0.10 1999 1999 2003 2003 4.0 4.0 4.0 4.0 5 0.40 1993 2002 2008 2008 12.1 12.5 6.5 15.5 Bolivia 8 0.70 1990 2001 2008 2008 12.6 13.5 7.5 18.5 Botswana 3 0.30 1990 2005 2008 2008 9.8 7.5 3.5 18.5 Brazil 83 7.20 1990 2007 1995 2008 7.8 7.5 0.5 18.5 Bulgaria 14 1.20 1997 2007 2002 2007 3.5 1.7 0.4 8.4 3 0.30 1998 2004 2007 2008 8.0 10.5 3.0 10.5 Cameroon 1 0.10 2004 2004 2008 2008 4.5 4.5 4.5 4.5 Chile 17 1.50 1990 2008 2001 2008 9.9 9.5 0.5 18.5 China 79 6.80 1994 2009 1998 2009 3.4 2.7 0.1 11.5 Colombia 25 2.20 1991 2008 2008 2008 5.6 4.5 0.5 17.5 Congo, Dem. Rep. 8 0.70 1996 2008 2007 2008 3.6 2.5 0.5 12.5 Costa Rica 5 0.40 1991 2008 2008 2008 10.5 9.5 0.5 17.5 Croatia 8 0.70 2000 2004 2003 2007 4.6 4.3 2.8 7.0 Czech Republic 14 1.20 1997 2005 2001 2008 4.1 4.9 0.4 8.1 Côte d'Ivoire 11 1.00 1990 2002 2008 2008 12.2 12.5 6.5 18.5 Dominican Republic 5 0.40 1990 2008 2008 2008 10.1 11.5 0.5 18.5 Ecuador 3 0.30 1999 2000 2008 2008 9.2 9.5 8.5 9.5 Egypt 2 0.20 2002 2006 2007 2007 3.1 3.1 1.3 5.0 El Salvador 6 0.50 1994 2005 2008 2008 8.5 8.5 3.5 14.5 Eritrea 1 0.10 1998 1998 2008 2008 10.5 10.5 10.5 10.5 Estonia 13 1.10 1996 2006 1997 2006 3.5 3.0 0.4 8.0 Gambia 2 0.20 1991 1991 2008 2008 17.5 17.5 17.5 17.5 Georgia 1 0.10 1997 1997 2007 2007 9.8 9.8 9.8 9.8 Ghana 10 0.90 1990 2008 2007 2008 10.2 7.3 0.5 18.5 Guatemala 3 0.30 1994 2000 2008 2008 10.5 8.5 8.5 14.5 Guinea 1 0.10 1996 1996 2008 2008 12.5 12.5 12.5 12.5 Guinea-Bissau 4 0.30 1990 2000 2008 2008 14.3 15.0 8.5 18.5 Guyana 2 0.20 2001 2006 2008 2008 5.0 5.0 2.5 7.5 Haiti 1 0.10 2000 2000 2008 2008 8.5 8.5 8.5 8.5 Honduras 2 0.20 1995 2008 2008 2008 7.0 7.0 0.5 13.5 Hong Kong 25 2.20 1973 1995 1975 2007 5.0 4.0 0.4 13.0 Hungary 10 0.90 1997 2004 1999 2007 3.9 3.5 0.9 9.1 India 59 5.10 1998 2007 1999 2008 3.3 2.3 0.2 9.0 Indonesia 51 4.40 1993 2005 2001 2007 3.7 3.4 0.2 13.3 Jamaica 3 0.30 1997 2004 2008 2008 7.5 6.5 4.5 11.5 Kazakhstan 1 0.10 1997 1997 2007 2007 10.3 10.3 10.3 10.3 Kenya 23 2.00 1990 2008 2007 2008 6.9 4.0 0.5 18.5

42

Korea, South 1 0.10 1993 1993 2003 2003 10.5 10.5 10.5 10.5 Kuwait 1 0.10 1998 1998 2007 2007 9.0 9.0 9.0 9.0 Latvia 6 0.50 2002 2005 2006 2006 2.2 2.0 1.0 4.0 Lithuania 12 1.00 1996 2005 1999 2006 3.3 2.5 1.0 6.0 Madagascar 4 0.30 1992 2008 2008 2008 8.3 8.0 0.5 16.5 Malawi 3 0.30 1995 2000 2008 2008 10.8 10.5 8.5 13.5 Malaysia 16 1.40 1995 2005 2001 2007 3.3 2.6 0.3 11.0 2 0.20 1995 1998 2008 2008 12.0 12.0 10.5 13.5 Mauritania 2 0.20 1991 1998 2008 2008 14.0 14.0 10.5 17.5 Mauritius 3 0.30 1990 1992 2008 2008 17.5 17.5 16.5 18.5 Mexico 45 3.90 1990 2008 1993 2008 7.9 7.5 0.4 18.5 Morocco 1 0.10 2008 2008 2008 2008 0.5 0.5 0.5 0.5 Mozambique 5 0.40 1998 2006 2007 2008 7.8 9.0 2.5 10.5 Namibia 2 0.20 1999 2002 2008 2008 8.0 8.0 6.5 9.5 Nicaragua 1 0.10 1999 1999 2008 2008 9.5 9.5 9.5 9.5 Nigeria 20 1.70 1993 2007 2007 2008 5.3 3.3 0.7 15.5 Panama 4 0.30 2000 2008 2008 2008 5.8 7.0 0.5 8.5 Peru 12 1.00 1994 2008 2008 2008 6.0 4.0 0.5 14.5 Philippines 7 0.60 1996 2005 2002 2008 6.1 5.7 3.3 10.4 Poland 123 10.60 1992 2006 1999 2008 7.0 6.3 0.1 13.8 Romania 23 2.00 1996 2005 2000 2006 5.3 5.4 1.0 10.0 Russian Federation 14 1.20 1995 2004 2002 2007 8.0 9.1 0.8 12.0 Rwanda 2 0.20 2004 2008 2007 2008 1.8 1.8 0.5 3.0 Saudi Arabia 1 0.10 1998 1998 2003 2003 5.4 5.4 5.4 5.4 5 0.40 1994 2004 2007 2008 9.0 9.5 3.0 14.5 Serbia 2 0.20 2003 2003 2007 2007 3.7 3.7 3.5 3.9 Singapore 2 0.20 1996 1997 2006 2007 10.4 10.4 10.2 10.7 Slovakia 7 0.60 2000 2005 2005 2006 3.9 4.3 0.5 5.3 Slovenia 2 0.20 2000 2000 2005 2005 4.9 4.9 4.9 4.9 South Africa 153 13.20 1990 2008 1993 2008 5.3 5.5 0.5 13.6 Sri Lanka 3 0.30 1997 2006 2004 2007 5.0 7.0 1.0 7.1 Swaziland 3 0.30 1990 2001 2008 2008 13.8 15.5 7.5 18.5 Taiwan 1 0.10 1996 1996 1998 1998 2.8 2.8 2.8 2.8 Tanzania 10 0.90 1994 2008 2008 2008 10.7 12.0 0.5 14.5 Thailand 36 3.10 1989 2006 1992 2008 2.1 1.1 0.2 6.4 4 0.30 1991 2001 2008 2008 12.8 13.0 7.5 17.5 Trinidad and Tobago 1 0.10 1990 1990 2008 2008 18.5 18.5 18.5 18.5 Tunisia 2 0.20 1997 2005 2007 2008 6.8 6.8 2.0 11.5 Turkey 2 0.20 2003 2006 2006 2008 2.6 2.6 2.0 3.3 Uganda 9 0.80 1993 2006 2007 2008 10.4 12.5 0.7 15.5 Ukraine 14 1.20 1996 2002 2004 2007 7.4 8.1 2.2 11.5 Uruguay 2 0.20 1990 2002 2008 2008 12.5 12.5 6.5 18.5 Venezuela 11 1.00 1991 2002 2008 2008 13.0 14.5 6.5 17.5 Zambia 6 0.50 1998 2008 2007 2008 7.8 9.5 0.5 10.5 Zimbabwe 16 1.40 1991 2002 2007 2008 11.7 11.5 4.7 17.5 Total/Mean/Min/Max 1,157 100.00 1973 2009 1975 2009 6.3 5.3 0.1 18.5

43

Table 2: Descriptive statistics of Gross IRRs from a USD investor’s perspective, winsorized at the 95th percentile, by country This table describes the distribution of the internal rates of return (IRR) that we observe from our sample transactions in the particular countries. The second column is the number of observations. The following present the mean, media, minimum, maximum and the standard deviation of the internal rates of return.

Country Obs. Mean Median Min Max Std. Dev. Algeria 1 0.6150.615 0.615 0.615 Angola 2 -0.005 -0.005 -0.034 0.023 0.041 Argentina 40 0.175 0.140 -0.501 1.480 0.427 Bangladesh 1 0.498 0.498 0.498 0.498 Benin 5 0.0710.092 -0.034 0.234 0.110 Bolivia 8 -0.062 -0.027 -0.500 0.290 0.230 Botswana 3 0.269 0.285 0.012 0.511 0.250 Brazil 83 0.0900.057 -1.000 1.480 0.683 Bulgaria 14 0.5150.245 -0.230 1.480 0.584 Burkina Faso 3 0.243 0.217 0.113 0.400 0.145 Cameroon 1 0.957 0.957 0.957 0.957 Chile 17 0.1590.132 -1.000 1.480 0.464 China 79 0.5610.325 -0.631 1.480 0.616 Colombia 25 0.222 0.127 -1.000 1.480 0.582 Congo, Dem. Rep. 8 0.168 0.085 -0.740 1.480 0.623 Costa Rica 5 -0.058 0.105 -0.823 0.272 0.445 Croatia 8 0.131 0.244 -0.348 0.414 0.281 Czech Republic 14 0.457 0.270 -0.210 1.480 0.502 Côte d'Ivoire 11 -0.196 0.067 -1.000 0.279 0.521 Dominican Republic 5 0.157 0.097 0.055 0.438 0.160 Ecuador 3 -0.616 -1.000 -1.000 0.151 0.665 Egypt 2 0.2950.295 0.178 0.412 0.165 El Salvador 6 0.355 0.332 -0.038 1.035 0.384 Eritrea 1 -0.204 -0.204 -0.204 -0.204 Estonia 13 0.2250.168 -0.641 1.480 0.495 Gambia 2 -0.499 -0.499 -1.000 0.001 0.708 Georgia 1 -0.129 -0.129 -0.129 -0.129 Ghana 10 0.1920.139 -0.255 0.744 0.305 Guatemala 3 -0.065 0.147 -0.806 0.465 0.661 Guinea 1 -1.000 -1.000 -1.000 -1.000 Guinea-Bissau 4 -0.345 -0.347 -1.000 0.312 0.606 Guyana 2 -0.078 -0.078 -0.190 0.035 0.159 Haiti 1 0.0000.000 0.000 0.000 Honduras 2 -0.038 -0.038 -0.376 0.300 0.478 Hong Kong 25 0.370 0.360 0.050 1.040 0.269 Hungary 10 0.3940.266 0.135 1.050 0.333 India 59 0.3970.300 -1.000 1.480 0.472 Indonesia 51 0.120 0.150 -1.000 0.831 0.348 Jamaica 3 0.9220.738 0.548 1.480 0.492 Kazakhstan 1 0.081 0.081 0.081 0.081 Kenya 23 0.0100.146 -1.000 0.527 0.443

44

Korea, South 1 0.021 0.021 0.021 0.021 Kuwait 1 0.5050.505 0.505 0.505 Latvia 6 0.4600.585 0.049 0.684 0.272 Lithuania 12 -0.196 -0.318 -1.000 1.095 0.619 Madagascar 4 0.096 0.091 -0.022 0.224 0.101 Malawi 3 0.027-0.092 -0.219 0.393 0.323 Malaysia 16 0.616 0.541 -0.160 1.480 0.463 Mali 2 0.1070.107 0.002 0.212 0.149 Mauritania 2 -0.071 -0.071 -0.443 0.301 0.526 Mauritius 3 -0.335 -0.064 -1.000 0.058 0.579 Mexico 45 0.0680.011 -1.000 1.480 0.541 Morocco 1 -0.438 -0.438 -0.438 -0.438 Mozambique 5 0.168 -0.010 -0.047 0.771 0.348 Namibia 2 0.2190.219 0.092 0.347 0.180 Nicaragua 1 0.104 0.104 0.104 0.104 Nigeria 20 0.2430.197 -1.000 1.480 0.614 Panama 4 0.0830.094 -0.009 0.152 0.070 Peru 12 0.0100.024 -1.000 1.480 0.659 Philippines 7 -0.082 -0.160 -1.000 0.819 0.548 Poland 123 0.2010.160 -1.000 1.480 0.497 Romania 23 0.3660.306 -0.230 1.420 0.389 Russian Federation 14 0.552 0.232 0.012 1.480 0.598 Rwanda 2 0.1140.114 -0.364 0.592 0.676 Saudi Arabia 1 0.115 0.115 0.115 0.115 Senegal 5 0.0870.039 -0.467 0.499 0.372 Serbia 2 0.1750.175 0.000 0.350 0.247 Singapore 2 -0.075 -0.075 -0.160 0.010 0.120 Slovakia 7 0.3260.367 -0.060 0.475 0.175 Slovenia 2 0.2150.215 0.130 0.300 0.120 South Africa 153 0.096 0.137 -1.000 1.480 0.548 Sri Lanka 3 0.207 0.080 0.013 0.527 0.279 Swaziland 3 -0.729 -1.000 -1.000 -0.187 0.469 Taiwan 1 0.4100.410 0.410 0.410 Tanzania 10 -0.019 -0.001 -0.353 0.366 0.251 Thailand 36 0.3340.265 -0.859 1.480 0.422 Togo 4 -0.253 -0.246 -1.000 0.478 0.654 Trinidad and Tobago 1 0.081 0.081 0.081 0.081 Tunisia 2 -0.056 -0.056 -0.142 0.031 0.122 Turkey 2 0.2710.271 0.213 0.330 0.083 Uganda 9 0.0020.132 -1.000 0.683 0.606 Ukraine 14 0.1790.151 0.036 0.529 0.141 Uruguay 2 -0.592 -0.592 -1.000 -0.185 0.576 Venezuela 11 -0.188 0.000 -1.000 0.221 0.426 Zambia 6 0.1340.050 -0.069 0.696 0.280 Zimbabwe 16 -0.188 -0.239 -1.000 1.480 0.651 Total/Mean/Min/Max 1,157 0.188 0.151 -1.000 1.480 0.536

45

Table 3: Industry segmentation of our sample This table presents the industries of the target companies of our sample transactions according to the Fama and French (1997) segmentation. I observe 14.1% of the 1,157 investee firms to be in the Trading industry, 10.8% to be in Communication, 7.4% to be in Business Services and so forth. All industry segments for which I observe less than 3.4% of the sample transactions are grouped as “Others” (including 11 transactions where no information on the investee firm’s industry is given). Altogether, I observe 47 different industries.

Fama and French Industry Classification % of observations Fin (Trading) 14.1% Telcm (Communication) 10.8% BusSv (Business Services) 7.4% Oil (Petroleum and Natural Gas) 6.1% Food (Food Products) 5.5% Rtail (Retail) 4.7% Trans (Transportation) 3.7% Hshld (Consumer Goods) 3.4% Others/unknown 44.3%

46

Table 4: Independent variables This table presents the independent variables, their dimensions, short descriptions and their sources. Dimen- Indicators sion Explanation Source(s) (1) Time Since the 1st Years Measure for the experience gained in general in a particular emerging Our data set Investment in the Host country. The first observed transaction in every country of our sample sets Country the cut-off date and receives a value of 0 for this variable. All other closing dates of the transactions in the same country (if any) use this offset. The years of experience are the difference between transaction closing and the first closing date. (2) GP's Experience in Years Similar to our principal variable “Time Since the 1st Investment in the Host Our data set the Host Country Country”, this experience measure counts the number of years between a subsequent investment which a GP might have and her first one in the same country. A GP’s first investment in a particular country sets the cut-off date for this GP in this country and receives a value of 0 for this variable. All subsequent investments (if any) by the same GP in the same country use this offset. The years of experience a GP gained in a particular country are the difference between transaction closing and the first closing date. (3) Time-Matching [%] Public market equivalent return which is measured by the geometric average Bloomberg S&P 500 Return growth rate of the S&P 500 index over the same time as the holding period of the PE transaction. (4) Time-Matching [%] Public market equivalent return of a local benchmark index. The calculation Bloomberg Local or Regional is the same as for indicator (3) but using local emerging stock market Stock Market Return indices instead of the S&P500 as benchmarks. For every particular country, the most important/representative stock market index is used. If such an index is not available, respectively was not available at the closing of the PE transaction, then a neighboring country or regional stock market index is used as alternative. A detailed list of the benchmark stock market indices and their alternatives is provided in Table 5. (5) Time-Matching [%] Public market equivalent from an emerging market local investor’s Bloomberg for the S&P S&P 500 Return in perspective. An investor headquartered in a particular emerging country has 500 and for the cross Local Currency the alternative to either invest in a PE transaction in her country or into the currency exchange S&P 500 index. To serve as a valid benchmark for this investor, the S&P rates 500 return needs to be corrected for fluctuations of the exchange rate between USD and the local currency over the same period as of the PE transaction. The calculation follows that of indicator (3) but corrects for cross currency rate changes. (6) Time-Matching [%] Follows the concept discussed for indicator (5) but uses the same local or Bloomberg for the Local or Regional regional stock market indices as for variable (4) as benchmarks. alternative benchmark Stock Market Return in indices and for the Local Currency cross currency exchange rates (7) Time-Matching [%] The transaction holding period-matching real GDP growth in the host International Monetary GDP Growth country of the investment. Corresponds to the geometric average growth rate Fund, International of a host country’s GDP between closing and exit of the PE transaction. The Financial Statistics and indicator’s accuracy is calculated on the level of months with the annual World Economic GDP observation broken down accordingly. Outlook/UN/national statistics (8) Aggregated IPO [billion This indicator aggregates the annual proceeds of IPO volumes (including Thomson One Banker Proceeds in Host USD] green shoe options) in a particular country for exit year of the PE Country in the Year of transaction. Exit (9) Host Country’s # The indicator is based on the Global Innovation Index of Cornell University, www.globalinnovationi Global Innovation INSEAD Business School and the World Intellectual Property Organization. ndex.org Index The Global Innovation Index relies on two sub-indices, the Innovation Input Sub-Index and the Innovation Output Sub-Index, each built around pillars. Five input pillars capture elements of the national economy that enable innovative activities: (1) Institutions, (2) Human capital and research, (3) Infrastructure, (4) Market sophistication, and (5) Business sophistication. Two output pillars capture actual evidence of innovation outputs: (6) Knowledge and technology outputs and (7) Creative outputs. Each pillar is divided into sub-pillars and each sub-pillar is composed of individual indicators (81 in total). Sub-pillar scores are calculated as the weighted average of individual indicators; pillar scores are calculated as the weighted average of sub-pillar scores. The index has been calculated in 2007 for the

47

Dimen- Indicators sion Explanation Source(s) first time. We use these 2007observations and regard them as time-invariant.

(10) Host Country’s # This data series measures the perceived quality of the educational system in World Economic Quality of the a country. The index ranges from 1 to 7, with higher values indicating that Forum Educational System the educational system in a country meets the needs of a competitive economy. Low values indicate that the system does not meet the needs of a competitive economy. We use the 2007 observations and regard them as time-invariant. (11) Host Country’s [%] Interest rate spread is the interest rate charged by banks on loans to prime World Economic Interest rate spread in customers minus the interest rate paid by commercial or similar banks for Forum; World the Year of Closing demand, time, or savings deposits. The observation year for this indicator Development matches the year of transaction closing unless the indicator was not yet Indicators calculated for that country. In these cases we refer to its first observation. (12) Difficulty of # The difficulty of firing index has eight components: World Bank (Doing Firing Index (i) whether redundancy is disallowed as a basis for terminating workers Business) (ii) whether the employer needs to notify a third party (such as a government agency) to terminate 1 redundant worker (iii) whether the employer needs to notify a third party to terminate a group of 25 redundant workers (iv) whether the employer needs approval from a third party to terminate 1 redundant worker (v) whether the employer needs approval from a third party to terminate a group of 25 redundant workers (vi) whether the law requires the employer to reassign or retrain a worker before making the worker redundant (vii) whether priority rules apply for redundancies (viii) whether priority rules apply for reemployment For the first question an answer of yes for workers of any income level gives a score of 10 and means that the rest of the questions do not apply. An answer of yes to question (iv) gives a score of 2. For every other question, if the answer is yes, a score of 1 is assigned; otherwise a score of 0 is given. Questions (i) and (iv), as the most restrictive regulations, have greater weight in the construction of the index. We regard this index as time- invariant and use the 2007 observations. (13) Host Country’s # The Property Rights Index is an assessment of the ability of individuals to Fraser Institute Property Rights Index accumulate private property, secured by clear laws that are fully enforced by the state. The index ranges from 1 to 10, with higher values indicating higher protection of property rights. We regard this index as time-invariant and use the 2007 observations. (14) GP’s Headquarter 0/1 Dummy variable equal to 1 if the headquarter of the GP that made the Our data set in the U.S. transaction is in the U.S. (15) Cross Border Deal 0/1 Dummy variable equal to 1 if the GP and the investee firm are not in the Our data set same country.

48

Table 5: Emerging Stock Market Benchmark Indices, respectively their Regional Benchmark Indices This table lists the benchmark indices I use to calculate the time-matching local stock market returns for the sample of PE transactions. The second column repeats the number of observations in our sample. The third one presents the local stock market index if any. The subsequent column counts the number of observations for which I can calculate a benchmark return using this index. If the index did not yet exist at the time of the respective PE transaction to benchmark or if the country does not have a representative stock market index at all, I refer to an alternative index from a neighboring country or a regional representative stock market. These alternatives are listed in the subsequent column. The final column shows the number of transactions for which I use the alternative benchmarks.

Country Obs. (total) Local Stock Obs. (local Regional Obs. (reg. Algeria 1 N/A 0 Morocco 1 Angola 2 N/A 0Nigeria 2 Argentina 40 MERVAL 40 0 Bangladesh 1 DSE 1 0 Benin 5 BRVM-Composite 1Morocco 4 Bolivia 8 N/A 0 Brazil 8 Botswana 3 BGSMDC 2 Kenya 1 Brazil 83 BOVESPA 83 0 Bulgaria 14 SOFIX 9 Turkey 5 Burkina Faso 3 BRVM-Composite 1 Morocco 2 Cameroon 1 N/A 0Nigeria 1 Chile 17 IPSA 17 0 China 79 SSE-A-Share 79 0 Colombia 25 IGBC A 24 Brazil 1 Congo, Dem. Rep. 8 N/A 0Kenya 8 Costa Rica 5 BTC 3 Brazil 2 Croatia 8 CROBEX 8 0 Czech Republic 14 PX 14 0 Côte d'Ivoire 11 BRVM-Composite 3 Morocco 8 Dominican Rep. 5 N/A 0 Brazil 5 Ecuador 3 ECU 3 0 Egypt 2 EGY30 2 0 El Salvador 6 N/A 0 Brazil 6 Eritrea 1 N/A 0Kenya 1 Estonia 13 OMX-TALLINN 13 0 Gambia 2 N/A 0 Morocco 2 Georgia 1 N/A 0 Turkey 1 Ghana 10 GSE-All-Share 5Morocco 5 Guatemala 3 N/A 0 Brazil 3 Guinea 1 N/A 0 Morocco 1 Guinea-Bissau 4 BRVM-Composite 1Morocco 3 Guyana 2 N/A 0 Brazil 2 Haiti 1 N/A 0 Brazil 1 Honduras 2 N/A 0 Brazil 2 Hong Kong 25 HSI 25 0 Hungary 10 BUX 10 0 India 59 SENSEX30 59 0 Indonesia 51 JCI-Composite 51 0 Jamaica 3 JSE 3 0

49

Country Obs. (total) Local Stock Obs. (local Regional Obs. (reg. Kazakhstan 1 KASE 0 Russian Federation 1 Kenya 23 NSE20 23 0 Korea, South 1 KOSPI200 1 0 Kuwait 1 KIC 1 0 Latvia 6 OMX-Riga 6 0 Lithuania 12 OMX-Vilnius B 12 0 Madagascar 4 N/A 0Kenya 4 Malawi 3 N/A 0Kenya 3 Malaysia 16 KLCI 16 0 Mali 2 BRVM-Composite 0Morocco 2 Mauritania 2 N/A 0 Morocco 2 Mauritius 3 SEMDEX 3 0 Mexico 45 IPC 45 0 Morocco 1 CFG25 1 0 Mozambique 5 N/A 0Kenya 5 Namibia 2 N/A 0Kenya 2 Nicaragua 1 N/A 0 Brazil 1 Nigeria 20 Nigeria-All-Share 16Kenya 4 Panama 4 BVPSI 4 0 Peru 12 IGBVL 12 0 Philippines 7 PSEi 7 0 Poland 123 WIG 123 0 Romania 23 BET 17 Turkey 6 Russian Federation 14 RTS C 14 0 Rwanda 2 N/A 0Kenya 2 Saudi Arabia 1 0 Kuwait 1 Senegal 5 BRVM-Composite 3Morocco 2 Serbia 2 BELEX-Line 0Turkey 2 Singapore 2 STI 0 Malaysia 2 Slovakia 7 SAX 7 0 Slovenia 2 LJSE 2 0 South Africa 153 JSE-All-Share 153 0 Sri Lanka 3 Colombo-All-Share 3 0 Swaziland 3 N/A 0Kenya 3 Taiwan 1 TAIEX 1 0 Tanzania 10 DSEI 1 Kenya 9 Thailand 36 SET 36 0 Togo 4 BRVM-Composite 2Morocco 2 Trinidad and 1 TTSE-Composite 0 Brazil 1 Tunisia 2 TUNINDEX 1Morocco 1 Turkey 2 XU100 2 0 Uganda 9 ALSI 1 Kenya 8 Ukraine 14 PFTS 4 Russian Federation 10 Uruguay 2 N/A 0Argentina 2 Venezuela 11 IBC 7 Peru 4 Zambia 6 LUSAKA 6 0 Zimbabwe 16 SE-Indus 0 Kenya 16 Total 1,157 987 170 A For values prior to July 2001 the Bogota Medellin General Index was rebased to match and extend the IGBC index B For values prior to December 1999 the Lithuania Litin G Index was rebased to match and extend the OMX Vilnius Index C For values prior to September 1995 the Russia RSF General Index was rebased to match and extend the Russia RTS index

50

Table 6: Descriptive statistics of the dependent and independent variables This table presents the mean, median, standard deviation, minimum, maximum values and the number of observations for the various dependent, independent and control variables. The relatively small mean and median values for the S&P 500 index return in USD and in local emerging country currencies, variables (3) and (5), are driven by the timing of the sample transactions. Their bulk has been closed around the year 2000 as presented in Figure 2. This was the period where the S&P 500 index peaked, leading to negative benchmark returns for transactions closed around the year 2000. The number of observations is reduced for variables (11) and (12) because the indicators are not available for several countries of our sample. Standard Variable Mean Median Deviation Min. Max. Obs. Winsorized IRR in USD 0.188 0.151 0.536 -1 1.480 1157 Winsorized IRR in Local Currency 0.218 0.177 0.549 -1 1.554 1157 (1) Time Since the 1st Investment in the 6.833 6.750 4.893 0 21.583 1157 Host Country (2) GP's Experience in the Host Country 3.844 2.000 4.466 0 21 1157 (3) Time-Matching S&P 500 Return 0.013 0.025 0.127 -0.502 0.505 1157 (4) Time-Matching Local or Regional 0.159 0.128 0.333 -0.841 3.380 1157 Stock Market Return (5) Time-Matching S&P 500 Return in 0.041 0.034 0.163 -0.860 0.978 1157 Local Currency (6) Time-Matching Local or Regional 0.173 0.147 0.272 -0.816 2.842 1157 Stock Market Return in Local Currency (7) Time-Matching GDP Growth 0.049 0.045 0.028 -0.063 0.192 1157 (8) Aggregated IPO Proceeds in Host 3.341 0.228 11.437 0 126.355 1157 Country in the Year of Exit (9) Host Country’s Global Innovation 2.716 2.770 0.488 1.530 4.412 1157 Index (10) Host Country’s Quality of the 3.453 3.7 0.741 2 6.2 1157 Educational System (11) Host Country’s Interest rate spread 0.096 0.056 0.102 0 0.669 1108 in the Year of Closing (12) Difficulty of Firing Index 30.695 30 23.234 0 100 1108 (13) Host Country’s Property Rights 4.519 4.634 1.428 1.322 7.970 1157 Index (14) GP’s Headquarter in the U.S. 0.499 0 0.501 0 1 1157 (15) Cross Border Deal 0.733 1 0.443 0 1 1157

51

Table 7: Correlation matrix Table 7 presents the correlation matrix among all independent and control variables. The two measures for country experience (1) and (2) are correlated at 0.61. However, they are not used simultaneously in any regression. The negative correlations between (3) and (1), respectively (2) indicate that I observe smaller benchmark returns for later transactions. This is plausible given the peak of the S&P 500 index in the year 2000. The time-matching emerging stock market returns do not follow this pattern and also their correlation with the S&P 500 is relatively small. Variable (3) is correlated with (5) and (4) with (6). This suggests that cross-currency effects do not strongly affect the results as benchmark returns do not strongly differ if calculated either in USD or in local emerging market currencies. Variables (7) and (8) are correlated, suggesting that IPOs happen after economic growth cycles. Additionally, (9) and (10) correlate. This is intuitive as one would expect better education yielding more innovations and vice versa. Further, (9) and (13) correlate which might be driven by the fact that innovations require legal protection. Pairs of variables with high correlations are either not used simultaneously or only at later stages of the stepwise regressions. (1) (2) (3) (4) (5) (6) (7) (8) Time Since the 1st Investment in the Host (1) 1.00 Country (2) GP's Experience in the Host Country 0.61 1.00 (3) Time-Matching S&P 500 Return -0.30 -0.42 1.00 Time-Matching Local or Regional Stock (4) 0.07 -0.15 0.34 1.00 Market Return Time-Matching S&P 500 Return in Local (5) -0.41 -0.31 0.75 -0.05 1.00 Currency Time-Matching Local or Regional Stock (6) -0.04 -0.17 0.46 0.84 0.28 1.00 Market Return in Local Currency (7) Time-Matching GDP Growth 0.10 -0.01 0.03 0.29 -0.21 0.21 1.00 Aggregated IPO Proceeds in the Host Country (8) 0.21 0.07 0.04 0.30 -0.08 0.28 0.51 1.00 in the Year of Exit (9) Host Country’s Global Innovation Index 0.14 -0.02 0.16 0.14 -0.00 0.09 0.29 0.32 Host Country’s Quality of the Educational (10) -0.09 -0.21 0.12 0.24 -0.05 0.18 0.22 0.14 System Host Country’s Interest Rate Spread in the (11) -0.03 0.13 -0.03 -0.12 0.14 -0.05 -0.17 -0.09 Year of Closing (12) Host Country’s Difficulty of Firing Index -0.14 -0.07 -0.07 0.12 -0.05 0.13 0.11 0.04 (13) Host Country’s Property Rights Index 0.29 0.13 0.11 -0.02 0.07 -0.00 0.04 0.17 (14) GP’s Headquarter in the U.S. -0.03 0.15 -0.39 -0.26 -0.19 -0.26 -0.19 -0.19 (15) Cross Border Deal -0.03 -0.05 -0.22 -0.01 -0.18 -0.03 0.09 -0.09

(9) (10) (11) (12) (13) (15) Host Country’s Quality of the Educational (10) 0.50 1.00 System Host Country’s Interest Rate Spread in the (11) -0.21 -0.33 1.00 Year of Closing (12) Host Country’s Difficulty of Firing Index 0.09 0.18 -0.25 1.00 (13) Host Country’s Property Rights Index 0.52 0.08 -0.06 -0.18 1.00 (14) GP’s Headquarter in the U.S. -0.33 -0.22 0.14 0.04 -0.30 1.00 (15) Cross Border Deal -0.19 0.03 -0.01 0.14 -0.30 0.60

52

Table 8 Panel A: PE transactions’ IRR determinants from a USD investor’s point of view This table presents OLS regression models A to E using various independent variables, controls and a constant. Dependent variable is always winsorized IRR of the PE transactions of the sample from a USD investor's point of view. Standard errors are robust. The first line with respect to every particular independent variable presents the estimated parameter coefficient. The second line presents the standardized coefficient (if all variables are transformed into z-scores), and the third line its standard error. In specification A, I regress the winsorized IRR in USD on the key variable of interest: “Time Since the 1st Investment in the Host Country”, fixed effects for countries, industries, and GPs, on the legal quality indicator and on a constant. The number of observations is 1157, and the adjusted R2 is 22.74%. The key variable of interest is significant at the 1% level and it remains significant at this level throughout all the specifications. Its economic impact on the IRR of a PE transaction can be interpreted as follows: Over the sampling horizon, one could expect an increase in the IRR of PE transactions by 1.3% from another year of waiting to enter, or equivalently, from another year of development of the emerging country, all else equal. Specifications B to E add independent variables. Model B introduces the PE transactions’ time matching return of the S&P 500 index. The parameter coefficient is significant at the 1% level and emphasizes the high correlation between the valuations in the private and public capital markets. Specification C analyzes the impact of emerging stock markets on the IRR of PE transactions additional to that of the S&P 500 index. The parameter is significant at the 1% level and reveals that the valuations of PE targets in emerging markets do not only follow the S&P 500 in general, but are superimposed by local stock market valuations. Model D adds the PE transactions’ time-matching GDP growth of the host country. Country fixed effects are dropped in this and the subsequent regression models because the GDP growth is partly captured in the country fixed effects (high versus low growth countries). Model E introduces the aggregated IPO proceeds in the host country for the year of the exit of the PE transaction to assess the liquidity of the local exit market. Both parameters, for GDP growth and IPO proceeds are significant at the 1% level, signaling the importance of economic growth and liquid exit markets for successful PE transaction making.

53

Specification: A B C D E β β β β β (Std.β) (Std.β) (Std.β) (Std.β) (Std.β) Independent Variables: [S.E.] [S.E.] [S.E.] [S.E.] [S.E.] Time Since the 1st 0.013*** 0.029*** 0.026*** 0.015*** 0.014*** Investment in the Host (0.119) (0.265) (0.235) (0.136) (0.125) Country [0.005] [0.006] [0.006] [0.004] [0.004]

Time-Matching S&P 1.150*** 0.980*** 0.729*** 0.744*** 500 Return (0.273) (0.232) (0.173) (0.176) [0.176] [0.185] [0.156] [0.153]

Time-Matching Local or 0.143*** 0.145*** 0.120** Regional Stock Market (0.089) (0.090) (0.074) Return [0.054] [0.053] [0.051]

Time-Matching GDP 2.509*** 2.003*** Growth (0.132) (0.105) [0.742] [0.760]

Aggregated IPO 0.006*** Proceeds in the Host (0.118) Country in the Year of [0.002] Exit

Constant 1.107*** 0.817*** 0.779*** 0.198 0.147 [0.211] [0.226] [0.229] [0.132] [0.135] Controls: Country Fixed Effects yes yes yes no no GP Fixed Effects yes yes yes yes yes Industry Fixed Effects yes yes yes yes yes Legal Quality yes yes yes yes yes N 1157 1157 1157 1157 1157 adj. R2 in % 22.74 26.55 26.94 22.66 23.18 P-values as of * p < 0.10, ** p < 0.05, *** p < 0.01

54

Table 8 Panel B: PE transactions’ IRR determinants from a USD investor’s point of View Panel B continues with the OLS regressions from Panel A. Dependent variable is always winsorized IRR of the PE transactions of our sample from a USD investor's point of view. Standard errors are robust. Specification F is equivalent to E from Panel A, but adds the time invariant indicator for the host countries’ capacities for innovation and replaces fund fixed effects with a dummy variable if the GP’s location is in the U.S. The model reveals that more innovative countries provide a better environment for successful PE transactions making. The coefficient is significant at the 1% level. The parameter of the GP location dummy is negative and significant at a 1% level. This highlights the notion that transactions structured from distant financial hubs perform worse than those where the GPs stay local. Specifications G and H replace ‘Time Since the 1st Investment in the Host Country’ by ‘GP’s Experience in the Host Country’ and introduce additional independent variables. The alternative experience indicator highlights the positive effect of learning. Its coefficient is significant at the 10% level. Its economic interpretation is as follows: Over the sample horizon, one could expect that another year of experience that a GP makes in a particular host country increases the IRR of a PE transaction by 0.8% on average, all else equal. Specifications G and H also reveal that, in addition to the innovation capacity of a country, its educational quality affects the outcome of PE transactions. Further, cost of debt, measured by the interest rate spread between inter-bank and prime corporate loans and cost of doing business, measured by the Difficulty of Firing Index impact the performance of PE transactions. The higher the financing cost at closing and the higher the general cost of doing business, the lower is the achieved IRR of PE transactions, all else equal. Both parameters are negatively significant at the 5% level at least.

55

Specification: F G H β β β (Std.β) (Std.β) (Std.β) Independent Variables: [S.E.] [S.E.] [S.E.] Time Since the 1st Investment in the Host Country 0.011*** (0.103) [0.003]

GP's Experience in the Host Country 0.007* 0.007* (0.060) (0.060) [0.004] [0.004]

Time-Matching S&P 500 Return 0.591*** 0.529*** 0.557*** (0.140) (0.125) (0.131) [0.142] [0.151] [0.158]

Time-Matching Local or Regional Stock Market Return 0.159*** 0.174*** 0.179*** (0.099) (0.108) (0.113) [0.050] [0.051] [0.053]

Time-Matching GDP Growth 1.594** 1.494** 1.336** (0.084) (0.078) (0.071) [0.656] [0.665] [0.671]

Aggregated IPO Proceeds in the Host Country in the Year 0.004*** 0.005*** 0.005*** of Exit (0.093) (0.109) (0.119) [0.002] [0.002] [0.002]

Host Country’s Global Innovation Index 0.135*** 0.095** 0.096** (0.123) (0.086) (0.077) [0.037] [0.043] [0.044]

Host Country’s Quality of the Educational System 0.048* 0.030 (0.066) (0.038) [0.026] [0.027]

Host Country’s Interest Rate Spread in the Year of Closing -0.004* (-0.067) [0.002]

Host Country’s Difficulty of Firing Index -0.002** (-0.069) [0.001]

GP’s Headquarter in the U.S. -0.110*** -0.111*** -0.104*** (-0.103) (-0.103) (-0.097) [0.038] [0.038] [0.039]

Constant 0.552*** 0.478*** 0.642*** [0.096] [0.114] [0.133] Controls: Country Fixed Effects no no no GP Fixed Effects no no no Industry Fixed Effects yes yes yes Legal Quality yes yes yes N 1157 1157 1108 adj. R2 in % 18.65 18.27 17.81 P-values as of * p < 0.10, ** p < 0.05, *** p < 0.01

56

Table 9: PE transactions’ IRR determinants from a local investor’s point of view This table presents OLS regression models I to L similar to the specifications in Table 8. However, the dependent variable is now replaced by the winsorized IRR of PE transactions from a local (host country) investor's perspective. Consequently, the IRRs are corrected for cross currency effects over the transactions’ holding periods. Standard errors are robust. The first line with respect to every particular independent variable presents the estimated coefficient. The second line presents the standardized coefficient, and the third line its standard error. In specification I, I regress the winsorized IRR in local currency on the key variable of interest, the time matching S&P 500 and the local stock market index return in the same (emerging country) currency, a constant, the legal quality indicator, country, fund, and industry fixed effects. In specification J, I add the transaction time matching GDP growth, the host country’s aggregated IPO proceeds in the exit year, its global innovation index score, and the three dummies for GP locations. At the same time, I drop the fund and the country fixed effects. In regressions K and L I switch to the alternative experience measure “GP’s Experience in the Host Country” and add the indicator for the educational quality (in K) and the difficulty of firing index (in L). As previously argued, including the difficulty of firing index in the regressions reduced the number of observations by 49. All results are qualitatively the same as in Table 8. However, the economic magnitude and statistical significance levels decrease slightly. Overall, the regressions provide evidence that the detected determinants of the IRR of emerging market PE transactions is not driven by cross currency effects between emerging market currencies and USD.

57

Specification: I J K L β β β β (Std.β) (Std.β) (Std.β) (Std.β) Independent Variables: [S.E.] [S.E.] [S.E.] [S.E.] Time Since the 1st Investment in the Host Country 0.028*** 0.013*** (0.247) (0.119) [0.006] [0.003]

GP's Experience in the Host Country 0.010** 0.009** (0.084) (0.077) [0.004] [0.004]

Time-Matching S&P 500 Return in Local Currency 0.898*** 0.735*** 0.660*** 0.660*** (0.266) (0.218) (0.195) (0.196) [0.155] [0.117] [0.115] [0.117]

Time-Matching Local or Regional Stock Market 0.131* 0.116* 0.126* 0.138* Return in Local Currency (0.065) (0.058) (0.062) (0.069) [0.079] [0.069] [0.071] [0.073]

Time-Matching GDP Growth 1.667** 1.511** 1.416* (0.085) (0.077) (0.073) [0.720] [0.731] [0.744]

Aggregated IPO Proceeds in Host Country in the 0.004*** 0.005*** 0.005*** Year of Exit (0.090) (0.107) (0.116) [0.002] [0.002] [0.002]

Host Country’s Global Innovation Index 0.135*** 0.097** 0.101** (0.120) (0.086) (0.079) [0.038] [0.045] [0.046]

Host Country’s Quality of the Educational System 0.046* 0.040 (0.062) (0.051) [0.028] [0.029]

Host Country’s Difficulty of Firing Index -0.001* (-0.055) [0.001]

GP’s Headquarter in the U.S. -0.099*** -0.103*** -0.099** (-0.090) (-0.094) (-0.090) [0.038] [0.038] [0.038]

Constant 0.771*** 0.533*** 0.481*** 0.557*** [0.230] [0.101] [0.119] [0.140] Controls: Country Fixed Effects yes no no no GP Fixed Effects yes no no no Industry Fixed Effects yes yes yes yes Legal Quality yes yes yes yes N 1157 1157 1157 1108 adj. R2 in % 25.36 16.45 16.08 15.42 P-values as of * p < 0.10, ** p < 0.05, *** p < 0.01

58

Table 10, Panel A: GP location and learning effects This table presents OLS regressions equivalent to those presented in Table 8, Panel A, but adding a dummy variable indicating cross border transactions and its interaction term with “Time Since the 1st Investment in the Host Country”. The dependent variable is always winsorized IRR of the PE transactions of the sample from a USD investor's point of view. Standard errors are robust. Specification M regresses the dependent variable on investee industry fixed effects, the legal quality indicator and the cross border deal dummy variable. The significant negative parameter of the dummy reveals that the IRRs of cross border deals are 11.6% below the transactions which are nationally originated. In specifications N to Q, I add independent variables. The results presented in specification N show that the learning effect remains if I control for cross border transactions. In specification O we add the interaction term between the first two variables. The cross border dummy itself loses significance but the negative parameter of the interaction term provides evidence that the benefit from learning diminishes in cross border transactions. Specifications P and Q support the robustness of the previous results. M N O P Q β β β β β (Std.β) (Std.β) (Std.β) (Std.β) (Std.β) [S.E.] [S.E.] [S.E.] [S.E.] [S.E.] Time Since the 1st 0.009*** 0.025*** 0.032*** 0.024*** Investment in the Host (0.085) (0.226) (0.293) (0.220) Country [0.003] [0.008] [0.008] [0.008]

Cross Border Deal -0.116*** -0.119*** 0.017 0.051 -0.003 (-0.096) (-0.098) (0.014) (0.042) (-0.002) [0.043] [0.043] [0.071] [0.070] [0.069]

Interaction of Time Since -0.019** -0.019** -0.014* First Investment and Cross (-0.191) (-0.185) (-0.136) Border Deal [0.009] [0.009] [0.008]

Time-Matching S&P 500 0.899*** 0.618*** Return (0.213) (0.147) [0.136] [0.146]

Time-Matching Local or 0.267*** Regional Stock Market (0.166) Return [0.057]

Constant 0.900*** 0.883*** 0.765*** 0.733*** 0.760*** [0.063] [0.063] [0.083] [0.082] [0.080] Country Fixed Effects no no no no no GP Fixed Effects no no no no no Industry Fixed Effects yes yes yes yes yes Legal Quality yes yes yes yes yes N 1157 1157 1157 1157 1157 adj. R2 in % 8.52 9.06 9.47 12.94 14.82 P-values as of * p < 0.10, ** p < 0.05, *** p < 0.01

59

Table 10, Panel B: GP location and learning effects This table presents OLS regressions equivalent to those presented in Table 8, Panel B, but adding a dummy variable indicating cross border transactions and its interaction term with “GP's Experience in the Host Country”. The dependent variable is always winsorized IRR of the PE transactions of the sample from a USD investor's point of view. Standard errors are robust. Regression specifications R to U reveal that the detrimental effect of not being local on learning remains with the alternative experience measure. The economic and statistical significance of the interaction term is even stronger than for the key variable “Times since 1st Investment in the Host Country”. R S T U β β β β (Std.β) (Std.β) (Std.β) (Std.β) [S.E.] [S.E.] [S.E.] [S.E.] GP's Experience in the Host Country 0.026*** 0.029*** 0.024*** 0.027*** (0.217) (0.246) (0.202) (0.225) [0.009] [0.009] [0.009] [0.009]

Cross Border Deal 0.024 0.039 -0.004 -0.020 (0.020) (0.032) (-0.004) (-0.016) [0.058] [0.059] [0.058] [0.056]

Interaction of GP's Experience in the -0.036*** -0.030*** -0.025** -0.029*** Host Country and Cross Border Deal (-0.294) (-0.246) (-0.201) (-0.232) [0.010] [0.010] [0.010] [0.010]

Time-Matching S&P 500 Return 0.708*** 0.443*** 0.470*** (0.168) (0.105) (0.111) [0.155] [0.163] [0.156]

Time-Matching Local or Regional 0.305*** 0.220*** Stock Market Return (0.189) (0.137) [0.059] [0.052]

Time-Matching GDP Growth 3.556*** (0.187) [0.625]

Constant 0.753*** 0.765*** 0.773*** 0.634*** [0.075] [0.075] [0.073] [0.077] Country Fixed Effects no no no no GP Fixed Effects no no no no Industry Fixed Effects yes yes yes yes Legal Quality yes yes yes yes N 1157 1157 1157 1157 adj. R2 in % 9.72 11.66 14.27 17.16 P-values as of * p < 0.10, ** p < 0.05, *** p < 0.01

60

Table 11: Robustness checks: Control for potential geographical bias This table presents OLS regressions equivalent to specification F of Table 8, Panel B, where always one particular country which might be over-weighted or very sensitive on the results is dropped. Most of the discussed parameters keep their economic and statistical significance. Specification: Without South Without Without Brazil Without China Africa Poland β β β β (Std.β) (Std.β) (Std.β) (Std.β) Independent Variables: [S.E.] [S.E.] [S.E.] [S.E.] Time Since the 1st Investment in the 0.014*** 0.011*** 0.009*** 0.010*** Host Country (0.132) (0.097) (0.083) (0.099) [0.003] [0.003] [0.003] [0.003]

Time-Matching S&P 500 Return 0.499*** 0.522*** 0.605*** 0.613*** (0.119) (0.126) (0.150) (0.150) [0.153] [0.147] [0.141] [0.143]

Time-Matching Local or Regional 0.198*** 0.179*** 0.124** 0.177*** Stock Market Return (0.130) (0.115) (0.079) (0.100) [0.053] [0.053] [0.049] [0.053]

Time-Matching GDP Growth 1.866*** 1.692** 1.543** 1.556** (0.103) (0.092) (0.085) (0.070) [0.680] [0.662] [0.648] [0.703]

Aggregated IPO Proceeds in the 0.003** 0.004** 0.004** 0.007 Host Country in the Year of Exit (0.075) (0.087) (0.083) (0.058) [0.002] [0.002] [0.002] [0.005]

Host Country’s Global Innovation 0.127*** 0.134*** 0.152*** 0.141*** Index (0.124) (0.127) (0.147) (0.132) [0.037] [0.037] [0.037] [0.040]

GP’s Headquarter in the U.S. -0.058 -0.116*** -0.104*** -0.123*** (-0.054) (-0.107) (-0.100) (-0.119) [0.040] [0.042] [0.039] [0.039]

Constant 0.429*** 0.550*** 0.522*** 0.574*** [0.098] [0.103] [0.099] [0.107] Controls: Country Fixed Effects no no no no GP Fixed Effects no no no no Industry Fixed Effects yes yes yes yes Legal Quality yes yes yes yes N 1004 1034 1074 1078 adj. R2 in % 20.57 19.94 18.33 15.30 P-values as of * p < 0.10, ** p < 0.05, *** p < 0.01

61

Table 12: Robustness checks: Impact of benchmark indices This table presents OLS regressions equivalent to specification F of Table 8, Panel B, where I control for the relatively high correlation among the two benchmark return measures and some other independent variables. Even if it is counterintuitive to assess PE returns without controlling for public market equivalents I drop the benchmark returns in these robustness checks. I first present specification F and then discard the benchmark returns individually in specifications V and W, and then both together in X. Specification: F V W X β β β β (Std.β) (Std.β) (Std.β) (Std.β) Independent Variables: [S.E.] [S.E.] [S.E.] [S.E.] Time Since the 1st Investment in the Host 0.011*** 0.006** 0.013*** 0.007** Country (0.103) (0.054) (0.120) (0.063) [0.003] [0.003] [0.003] [0.003]

Time-Matching S&P 500 Return 0.591*** 0.739*** (0.140) (0.175) [0.142] [0.134]

Time-Matching Local or Regional Stock 0.159*** 0.236*** Market Return (0.099) (0.146) [0.050] [0.055]

Time-Matching GDP Growth 1.594** 1.430** 1.880*** 1.840*** (0.084) (0.075) (0.099) (0.097) [0.656] [0.655] [0.653] [0.665]

Aggregated IPO Proceeds in the Host 0.004*** 0.004*** 0.005*** 0.005*** Country in the Year of Exit (0.093) (0.088) (0.108) (0.110) [0.002] [0.002] [0.002] [0.002]

Host Country’s Global Innovation Index 0.135*** 0.139*** 0.133*** 0.138*** (0.123) (0.127) (0.121) (0.126) [0.037] [0.036] [0.037] [0.037]

GP’s Headquarter in the U.S. -0.110*** -0.150*** -0.117*** -0.179*** (-0.103) (-0.140) (-0.109) (-0.167) [0.038] [0.037] [0.038] [0.037]

Constant 0.552*** 0.585*** 0.568*** 0.627*** [0.096] [0.095] [0.097] [0.095] Controls: Country Fixed Effects no no no no GP Fixed Effects no no no no Industry Fixed Effects yes yes yes yes Legal Quality yes yes yes yes N 1157 1157 1157 1157 adj. R2 in % 18.65 17.48 18.07 15.93 P-values as of * p < 0.10, ** p < 0.05, *** p < 0.01

62

Table 13: Robustness checks: Impact of the cut-off date for the experience measure and of the transactions prior to 1988 This table presents OLS regressions equivalent to those presented in Table 8, Panel A. The dependent variable is always winsorized IRR of the PE transactions of our sample from a USD investor's point of view. Standard errors are robust. Regression specifications Y to AC address a bias eventually caused by not observing the pioneering transactions in each country of the sample. I drop the first five transactions and recalculate our experience measure using the closing date of sixth transaction as new cut-off for the calculation of the experience measure. This reduces the sample to 856 observations. The final column (>1987) presents the regression where 17 transactions prior to 1988 are dropped. The Table provides evidence that all discussed results hold and are not effect by outliers prior to 1988 or by not observing the “real pioneering” transactions. Y Z AA AB AC >1987 β β β β β β (Std.β) (Std.β) (Std.β) (Std.β) (Std.β) (Std.β) [S.E.] [S.E.] [S.E.] [S.E.] [S.E.] [S.E.] Time Since the 1st Investment 0.030*** 0.044*** 0.041*** 0.014** 0.011* 0.015*** in the Host Country (0.189) (0.283) (0.259) (0.089) (0.071) (0.133) [0.008] [0.009] [0.008] [0.006] [0.006] [0.004]

Time-Matching S&P 500 1.164*** 1.026*** 0.649*** 0.681*** 0.750*** Return (0.259) (0.228) (0.144) (0.151) (0.176) [0.220] [0.225] [0.194] [0.190] [0.159]

Time-Matching Local or 0.130** 0.166** 0.136** 0.133** Regional Stock Market (0.080) (0.102) (0.083) (0.081) Return [0.059] [0.065] [0.063] [0.053]

Time-Matching GDP Growth 3.623*** 2.803** 1.889** (0.176) (0.136) (0.097) [1.081] [1.124] [0.773]

Aggregated IPO Proceeds in 0.007*** 0.006*** the Host Country in the Year (0.143) (0.119) of Exit [0.002] [0.002]

Constant 1.030*** 0.695** 0.671** 0.236 0.203 0.136 [0.260] [0.286] [0.290] [0.185] [0.188] [0.138] Country Fixed Effects yes yes yes no no no GP Fixed Effects yes yes yes yes yes yes Industry Fixed Effects yes yes yes yes yes yes Legal Quality yes yes yes yes yes yes N 856 856 856 856 856 1140 adj. R2 in % 24.93 28.20 28.46 22.72 23.46 23.28 P-values as of * p < 0.10, ** p < 0.05, *** p < 0.01

63

Table 14: Decreasing marginal benefit of waiting This table presents level-log regressions of the PE transactions’ winsorized IRR in USD on several independent variables. I add 1 year to our key variable “Time Since the 1st Investment in the Host Country” to allow the logarithmic transformation for all observations. Standard errors are robust. The first line with respect to every particular independent variable presents the estimated parameter coefficient. The second line shows the standardized coefficient and the third line its standard error. Using the natural logarithm of the experience measure takes the expectation of its decreasing marginal effect into account. All the above discussed results hold and the significance of the key variable of interest throughout all the regressions suggests that the experience has a non-linear decreasing marginal effect on achieved IRRs.

64

Specification: AD AE AF β β β (Std.β) (Std.β) (Std.β) Independent Variables: [S.E.] [S.E.] [S.E.] log(Time Since the 1st Investment in the Host 0.055*** 0.063*** 0.058*** Country+1) (0.090) (0.104) (0.094) [0.020] [0.018] [0.018]

Time-Matching S&P 500 Return 0.606*** 0.576*** 0.590*** (0.144) (0.137) (0.138) [0.149] [0.136] [0.143]

Time-Matching Local or Regional Stock Market 0.162*** 0.146*** 0.153*** Return (0.101) (0.090) (0.096) [0.056] [0.052] [0.054]

Time-Matching GDP Growth 2.483*** 1.498** 1.356** (0.130) (0.079) (0.072) [0.740] [0.654] [0.662]

Aggregated IPO Proceeds in the Host Country in 0.005*** 0.005*** the Year of Exit (0.105) (0.114) [0.002] [0.002]

Host Country’s Global Innovation Index 0.083* 0.086** (0.076) (0.069) [0.043] [0.044]

Host Country’s Quality of the Educational 0.052** 0.035 System (0.072) (0.046) [0.026] [0.027]

Host Country’s Interest Rate Spread in the Year -0.003 of Closing (-0.057) [0.002]

Host Country’s Difficulty of Firing Index -0.001** (-0.061) [0.001]

GP’s Headquarter in the U.S. -0.108*** -0.102*** (-0.100) (-0.096) [0.038] [0.039]

Constant 0.198 0.390*** 0.541*** [0.131] [0.115] [0.135] Controls: Country Fixed Effects no no no GP Fixed Effects yes no no Industry Fixed Effects yes yes yes Legal Quality yes yes yes N 1157 1157 1108 adj. R2 in % 22.12 18.85 18.24 P-values as of * p < 0.10, ** p < 0.05, *** p < 0.01

65

Figure 1: Frequency of closing years of sample transactions 200 150 100 Frequency 50 0 1970 1980 1990 2000 2010 Year of Transaction Closing

66

Figure 2: Distribution of winsorized IRRs in USD 150 100 Frequency 50 0 -1 -.5 0 .5 1 1.5 Gross IRR USD - Winsorized at the 95 Percentile

67